A low-altitude target radar track space-time fusion discrimination method

Abstract

The application relates to the technical field of low-altitude monitoring, and discloses a low-altitude target radar track space-time fusion discrimination method, which comprises the following steps: a data processing computer accesses radar data, electronic reconnaissance data and photoelectric data, field unification and space coordinate unification are carried out on the source data, a track position point in a radar track point sequence is generated based on a ranging value, an azimuth angle and a pitch angle output by a radar device; a chord deviation stability value is generated based on a midpoint chord deviation, a chord deviation change rate and a same direction degree of front and rear sections in a five-point window; a fusion reference time is determined, the electronic reconnaissance data and the photoelectric data are time-aligned, and a same target candidate set is formed according to a space interval and a correlation threshold; a radar unmanned aerial vehicle support degree and a radar category label are determined according to the chord deviation stability value, a radar voting weight is adjusted, a fusion category is obtained by combining an electronic reconnaissance category label and a photoelectric category label, and a unified target track, the fusion category and a category support degree are output.

Description

Technical Field

[0001] This invention relates to the field of low-altitude surveillance technology, and more specifically, to a spatiotemporal fusion discrimination method for low-altitude target radar tracks. Background Technology

[0002] Low-altitude surveillance systems typically integrate radar, electronic reconnaissance, and optoelectronic equipment. Radar equipment continuously outputs range, azimuth, elevation, track number, and sampling time; electronic reconnaissance equipment outputs radiation source-related records and azimuth lines; and optoelectronic equipment outputs target frames, line-of-sight angles, and category markers. This data, after entering the data processing computer via network interface, device message bus, or serial interface, needs to be converted to a unified format before entering the target association, track maintenance, and monitoring display processes.

[0003] In actual operation, the refresh cycles, timestamp sources, coordinate references, measurement errors, and quality status of each device are inconsistent. Radar track recording is more suitable for continuous tracking, electronic reconnaissance recording can provide information related to radiation sources, and photoelectric recording can provide image tracking results. However, any device may experience recording gaps due to obstruction, low signal-to-noise ratio, limited field of view, or short-term missed detections. The data processing computer needs to select available records near the same fusion reference time and complete spatial correlation under a unified coordinate system in order to classify the outputs of different devices into the same target candidate set.

[0004] In existing low-altitude surveillance data processing workflows, radar systems often rely on changes in heading, velocity, or fitting residuals for category determination. These individual indicators are prone to fluctuation when target distance changes, point-to-point spacing changes, short-term missed points, and maneuvering occur simultaneously. If radar category results are still entered into multi-source voting with fixed weights, and electronic reconnaissance category markers and electro-optical category markers are in a state of missing data or low quality, then unnecessary jumps in the fused category at adjacent fusion reference times may occur.

[0005] Therefore, without changing the radar equipment's acquisition method, the low-altitude surveillance data processing flow needs a local continuity evaluation method for radar track point sequences. This evaluation result should be passed to the radar voting weight and multi-source category support generation process, so that the unified target track, fused category, and category support can be updated synchronously with the equipment data quality and local track changes, making it easier for the surveillance terminal to stably display the fusion result of the same low-altitude target. Summary of the Invention

[0006] This invention provides a spatiotemporal fusion discrimination method for low-altitude target radar tracks, which solves the technical problems mentioned in the background art.

[0007] To achieve the above-mentioned objectives, the present invention provides the following technical solution.

[0008] Compared with the prior art, the present invention has the following substantial features and significant progress: This invention integrates radar, electronic reconnaissance, and optoelectronic equipment data into a data processing computer. By unifying fields, coordinates, multi-source time alignment, and spatial correlation, records from different devices are grouped into the same target candidate set. Then, a stable chordal deviation value is generated using the midpoint chordal deviation, chordal deviation rate of change, and unidirectionality of preceding and following segments in the radar track point sequence. Based on this, radar UAV support, radar category labels, and radar voting weights are formed, ensuring that radar output is no longer solely based on a single turning angle or residual for voting. Category support is generated with the participation of electronic reconnaissance and optoelectronic category labels, and continuous confirmation is performed using the preceding fusion reference time. This reduces the impact of short-term omissions, point spacing changes, and device label fluctuations on the fused category output within the data reception, processing, storage, and display chain, improving the consistency of unified target track maintenance and monitoring terminal display. Attached Figure Description

[0009] Figure 1 This is a system architecture and processing mechanism diagram of a spatiotemporal fusion discrimination method for low-altitude target radar tracks according to the present invention. Detailed Implementation

[0010] The following description is provided in conjunction with the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the invention; those skilled in the art can make equivalent substitutions or combinations for the specific implementations.

[0011] Example 1: This embodiment provides a spatiotemporal fusion discrimination method for low-altitude target radar tracks, applied to a data processing computer. The data processing computer receives radar data, electronic reconnaissance data, and electro-optical data, and performs field unification and spatial coordinate unification on each source data. Then, it generates chord-biased stable values ​​using local geometric relationships in the radar track point sequence. Based on these chord-biased stable values, it determines the radar UAV support and radar category label. Finally, it combines the electronic reconnaissance category label and the electro-optical category label to obtain the fused category, category support, and unified target track. This embodiment is designed for continuous monitoring of UAVs, birds, or unknown low-altitude targets within a low-altitude surveillance area.

[0012] like Figure 1 As shown, radar equipment 100, electronic reconnaissance equipment 200, and optoelectronic equipment 300 respectively send radar data, electronic reconnaissance data, and optoelectronic data to data processing computer 400. Data processing computer 400 is equipped with a field coordinate unification unit 410, a chord deflection stability value generation unit 420, and a spatiotemporal correlation fusion discrimination unit 430. Monitoring and display terminal 500 receives the unified target track, fusion category, and category support output by data processing computer 400.

[0013] Figure 1 In this process, radar device 100, electronic reconnaissance device 200, and optoelectronic device 300 input multi-source data to data processing computer 400 via a left-hand one-way arrow. This one-way arrow indicates the data processing flow of the device entering the data processing computer 400, but does not indicate that the data processing computer 400 reversely controls radar device 100, electronic reconnaissance device 200, or optoelectronic device 300. Field coordinate unification unit 410 sends the unified radar track point sequence to chord-biased stability value generation unit 420 via a downward arrow. Chord-biased stability value generation unit 420 sends chord-biased stability values, radar UAV support, and radar category labels to spatiotemporal correlation fusion discrimination unit 430 via a downward arrow. Field coordinate unification unit 410 also sends unified electronic reconnaissance data and optoelectronic data to spatiotemporal correlation fusion discrimination unit 430 via a branch arrow, enabling the electronic reconnaissance data and optoelectronic data to participate in multi-source time alignment, spatial correlation, and fusion category generation. Spatiotemporal correlation fusion discrimination unit 430 sends unified target tracks, fusion categories, and category support to monitoring and display terminal 500 via a one-way arrow.

[0014] In this embodiment, the data processing computer can be a server, edge computing device, embedded processing unit, or task processing node in a low-altitude surveillance system.

[0015] The radar data processed in this embodiment includes at least track number, sampling time, range value, azimuth angle, elevation angle, radial velocity, echo quality, and track status. Electronic reconnaissance data includes at least radiation source markers, direction finding results, frequency band information, sampling time, and electronic reconnaissance category markers. Electro-optical data includes at least target frame, line-of-sight angle, sampling time, tracking status, and electro-optical category markers.

[0016] Example 2: In one implementation, the data processing computer 400 accesses radar data, electronic reconnaissance data, and optoelectronic data according to their device sources. The radar data transmitted by the radar device 100 serves as the basis for generating radar track point sequences, track position points, and chord deflection stability values; the electronic reconnaissance data transmitted by the electronic reconnaissance device 200 serves as an auxiliary category source in multi-source time alignment and fusion category generation; and the optoelectronic data transmitted by the optoelectronic device 300 serves as another auxiliary category source in multi-source time alignment, spatial correlation, and fusion category generation.

[0017] The data processing computer 400 writes the device number and reception time into each piece of data upon receiving it, and distinguishes radar data, electronic reconnaissance data, and electro-optical data according to their device source. The device number is used to differentiate between different device sources during subsequent cleaning and sorting, preventing records from different devices from being mistakenly identified as records from the same source in the same processing flow. The reception time is used to determine the retention order of duplicate records when quality indicators are the same, ensuring that duplicate records from the same device, the same track number, and the same sampling time can form deterministic processing results.

[0018] Track numbers in radar data are used to identify a sequence of radar track points. The sampling time indicates when radar device 100 generated the radar data. Range, azimuth, and elevation angles are used to generate track position points. Radial velocity can serve as one source of velocity information in a unified target track. Echo quality and track status are used to determine whether radar data is suitable for inclusion in the same target candidate set.

[0019] Radiation source markers in electronic reconnaissance data are used to identify target-related radiation sources detected by electronic reconnaissance equipment 200. Direction finding results are used to form electronic reconnaissance azimuth lines. Frequency band information is used to assist in confirming the validity of the electronic reconnaissance data. Sampling time is used for comparison with a fusion reference time. Electronic reconnaissance category markers are used for fusion category generation. Target bounding boxes in optoelectronic data are used to identify target areas detected by optoelectronic equipment 300. The line-of-sight angle is used to form the optical television axis azimuth. Sampling time is used for comparison with a fusion reference time. Tracking status is used to determine the validity of the optoelectronic data. Optoelectronic category markers are used for fusion category generation.

[0020] When the data processing computer 400 receives data from multiple devices simultaneously, the receiving order is not the sole basis for the target association result. The data processing computer 400 uses the device number, track number, and sampling time as the basic fields for subsequent cleaning and sorting, and uses the receiving time as an auxiliary basis for the order in which duplicate records are retained. For records within the same device source, the sampling time is used first to restore the device acquisition order; for records from different device sources, the fusion reference time is used to unify the multi-source time alignment benchmark.

[0021] Example 3: The field coordinate unification unit 410 performs field unification and spatial coordinate unification. Field unification includes unifying distance units to meters, angle units to radians, and time units to milliseconds. If radar device 100, electronic reconnaissance device 200, or optoelectronic device 300 outputs data with different field names or different units, the field coordinate unification unit 410 first reads the original fields according to the device data dictionary, and then generates data in a unified format. The unified format data retains both the original field content and the unified field content, allowing subsequent processing to trace the data source.

[0022] During the field standardization process, the range value is the slant distance from the radar station to the target point, standardized in meters. The azimuth angle is the horizontal rotation angle of the target relative to the radar station, standardized in radians. The elevation angle is the vertical rotation angle of the target relative to the radar station, standardized in radians. The sampling time is the time when the device generated the data or the device time that can represent the data, standardized in milliseconds. The time tolerance is the maximum time difference that allows records from different devices to be included in the same time slice, also expressed in milliseconds.

[0023] The spatial coordinates are unified with the radar station or monitoring area reference point as the origin. The range, azimuth, and elevation angles output by the radar equipment 100 are converted into eastward, northward, and altitude positions, and the track position points are composed of the eastward, northward, and altitude positions. The eastward, northward, and altitude positions are all position components in the unified coordinate system.

[0024]

[0025] in, Indicates the eastward position. Indicates the distance measurement value. Indicates pitch angle, This indicates the azimuth angle. The unit for eastward position is meters, the unit for distance measurement is meters, and the units for elevation and azimuth angles are radians.

[0026]

[0027] in, Indicates the northward position. Indicates the distance measurement value. Indicates pitch angle, This indicates the azimuth angle. The unit for northward position is meters, the unit for distance measurement is meters, and the units for elevation and azimuth angles are radians.

[0028]

[0029] in, Indicates altitude position. Indicates the distance measurement value. This indicates the pitch angle. The unit of measurement for altitude is meters, the unit of measurement for distance is meters, and the unit of measurement for pitch angle is radians.

[0030]

[0031] in, Indicates the position point of the flight path. Indicates the eastward position. Indicates the northward position. This indicates altitude. The track position point is a three-dimensional position vector with the dimension of meters.

[0032] For device data that has already been output in rectangular coordinates, the field coordinate unification unit 410 converts it to a unified coordinate system according to the device installation position, orientation calibration value, and coordinate transformation matrix. The device installation position is used to determine the translation relationship of the device coordinate system relative to the unified coordinate system, the orientation calibration value is used to determine the rotation relationship of the device coordinate system relative to the unified coordinate system, and the coordinate transformation matrix is ​​used to convert the position vector in the device coordinate system to the position vector in the unified coordinate system.

[0033] When the device provides velocity components, the field coordinate unification unit 410 can convert the velocity components into eastward velocity, northward velocity, and vertical velocity. When the device does not provide velocity components but continuous track position points exist, the velocity components can be calculated based on the time difference between adjacent track position points and their corresponding sampling times. The eastward velocity, northward velocity, and vertical velocity are used to generate a unified target track and are not required inputs for generating chord-deflection stable values.

[0034] Example 4: Before generating the chord-biased stability value, the chord-biased stability value generation unit 420 cleans and sorts the radar track point sequence. The purpose of cleaning is to ensure that the radar data entering the five-point window has the necessary fields and a reasonable measurement range, and the purpose of sorting is to maintain the sampling order of the radar track point sequence. Cleaning and sorting do not change the number of track points, the connection relationship between points, or the sampling order already formed by the radar device 100; they only exclude records that cannot be used for correlation or calculation.

[0035] During the cleaning process, records lacking track numbers, sampling times, range values, azimuth angles, or elevation angles are marked as unusable for association. When a track number is missing, the data processing computer 400 cannot determine the radar track point sequence to which the record belongs; when a sampling time is missing, the data processing computer 400 cannot perform multi-source time alignment and sorting; when range values, azimuth angles, or elevation angles are missing, the data processing computer 400 cannot generate track position points. Therefore, these records are not included in the formation of the same target candidate set, nor are they used for chord deflection stability value calculation.

[0036] During the cleaning process, records exceeding the equipment's measurement range or angular range are marked as anomalous records. The equipment range is determined by the engineering configuration of radar equipment 100, and the angular range is determined by the range of azimuth and elevation angles that radar equipment 100 can output. For records containing a velocity field, records exceeding the velocity range are marked as anomalous records. The velocity range is determined by the engineering configuration of radar equipment 100 or the effective target velocity range of the low-altitude surveillance system. Anomalous records are excluded from the formation of the same target candidate set to avoid obviously invalid radar data affecting the fusion category.

[0037] The measurement range, angle range, and velocity range are read from the radar device 100's device data dictionary, device configuration file, or device calibration record. If multiple range values ​​exist in the device data dictionary, device configuration file, and device calibration record, the device calibration record takes precedence, followed by the device configuration file, with the device data dictionary serving as a fallback. If no corresponding range value exists in any of the three sources, this field is not included in the anomaly record determination, but the record is written into the data quality status and its quality reduction factor is lowered in the subsequent current voting weights.

[0038] During sorting, the data processing computer 400 sorts the data according to the device number, track number, and sampling time. The device number is used to distinguish different radar devices or different sources, the track number is used to distinguish different radar track point sequences generated by the same device, and the sampling time is used to reconstruct the temporal order within the same radar track point sequence. The sorted radar track point sequence serves as the basis for establishing a continuous five-point window.

[0039] For duplicate records from the same device, track number, and sampling time, the data processing computer 400 retains the record with the higher quality flag. If the quality flags are the same, the record with the earlier reception time is retained. The quality flag comes from the echo quality or track status output by the radar device 100. This duplicate record processing rule ensures that at most one valid track position point is retained in the same radar track point sequence for each sampling time, avoiding repeated participation in the calculation of adjacent point spacing, midpoint chord offset, and chord offset change rate at the same sampling time.

[0040] Quality indicators are used in the order of repeating records according to three levels: valid, downgraded, and invalid. Valid level is higher than downgraded level, and downgraded level is higher than invalid level. If both echo quality and track status are present in the radar equipment 100 output, the quality indicator level is first determined by the track status, and then the echo quality is used to sort within the same quality indicator level. If a quality indicator is missing, the record is processed as downgraded. Radar equipment 100 is mapped as valid when its track status is stable tracking and the echo quality is not lower than the equipment's effective quality limit; it is mapped as downgraded when its track status is transient tracking or the echo quality is lower than the equipment's effective quality limit but not lower than the equipment's minimum quality limit; and it is mapped as invalid when its track status is lost, deleted, or the echo quality is lower than the equipment's minimum quality limit. Electronic reconnaissance equipment 200 is mapped as valid when its frequency band information is identifiable, direction finding results are present, and the equipment alarm status is normal; it is mapped as downgraded when frequency band information is missing but direction finding results are present, or when the equipment alarm status is downgraded; and it is mapped as invalid when direction finding results are missing, radiation source markers are invalid, or the equipment alarm status is invalid. When the tracking status of the optoelectronic device 300 is stable tracking and the target frame is valid, it is mapped to valid. When the tracking status is transient tracking or short-term loss but the target frame can still be traced back, it is mapped to downgraded. When the target frame is missing, the tracking status is lost, or the device alarm status is invalid, it is mapped to invalid.

[0041] When the sampling time difference between two adjacent valid records exceeds the preset breakpoint duration, the data processing computer 400 can split the original radar track point sequence into two radar track point sequence segments. The preset breakpoint duration is determined based on the sampling period of the radar device 100 and the allowable short-term missing time for low-altitude targets. This processing is only used to avoid five consecutive point windows spanning discontinuous track segments due to long-term missing data.

[0042] Example 5: The string deflection stability value generation unit 420 establishes a continuous five-point window in the cleaned and sorted radar track point sequence. Within this continuous five-point window, the current track point, its two preceding track points, and its two following track points are determined. If the current track point is the... The first track position point, then a five-point window contains the first... The first track position point, the first The first track position point, the first The first track position point, the first The first track position and the first Each track position point. Five consecutive point windows slide only along the original sampling order of the radar equipment 100, without reordering or inserting virtual track points.

[0043] Entering and leaving line segment vectors are generated based on adjacent track position points. The entering line segment vector represents a three-dimensional vector pointing from the previous track point to the current track point, and the leaving line segment vector represents a three-dimensional vector pointing from the current track point to the next track point. Both entering and leaving line segment vectors are obtained by subtracting the track position points, and their units are meters.

[0044]

[0045] in, This indicates entering the line segment vector. This indicates the track position point corresponding to the current track point. This indicates the track position point corresponding to the previous track point.

[0046]

[0047] in, Indicates the vector leaving the line segment. This indicates the track position point corresponding to the next track point. This indicates the track position point corresponding to the current track point.

[0048] The following parameters are generated based on the entry and exit vectors: cross-point distance, adjacent-point spacing, midpoint chord offset, and forward / backward segment homogeneity. Cross-point distance represents the three-dimensional distance from one track point to the next. Adjacent-point spacing represents the three-dimensional distance between two adjacent radar track points. Midpoint chord offset represents the perpendicular distance from the intermediate track point to the line connecting the preceding and following points. Forward / backward segment homogeneity indicates the degree of consistency in the directions of the preceding and following track segments.

[0049]

[0050] in, Indicates the distance across points. This indicates entering the line segment vector. This represents the vector leaving the line segment. The distance across the point is measured in meters.

[0051]

[0052] in, Indicates the distance between adjacent points. This represents the vector entering the line segment. The distance between adjacent points is measured in meters.

[0053]

[0054] in, Indicates the distance between adjacent points. This represents the vector that leaves the line segment.

[0055]

[0056] in, Indicates the deflection of the midpoint chord. This indicates entering the line segment vector. Indicates the vector leaving the line segment. Indicates the distance across points. This indicates a minimum length that is not divisible by zero. The unit of measurement for the midpoint chord deflection is meters, and the unit of measurement for the minimum length that is not divisible by zero is used to prevent division by zero due to excessively small distances between points.

[0057]

[0058] in, Indicates the same directionality between the preceding and following segments. This indicates entering the line segment vector. Indicates the vector leaving the line segment. and Indicates the distance between adjacent points. This indicates a decimal portion of area. The unit of measurement for this decimal portion of area is square meters, and it is used in conjunction with... Maintain consistent dimensions. The unidirectionality of the preceding and following segments is a dimensionless value; it approaches near-uniformity when the directions of the two preceding and following track segments are essentially the same. It decreases when a significant reversal occurs.

[0059] The chord deviation ratio is generated based on the distance between adjacent points and the midpoint chord deviation. The chord deviation ratio represents the proportion of the midpoint chord deviation relative to the distance between adjacent points, and is used to reduce the impact of changes in target distance and inter-point distance on the absolute deviation scale. The larger the chord deviation ratio, the more significant the local deviation of the current track from the line connecting the two points before and after.

[0060]

[0061] in, Indicates the chord deflection ratio. Indicates the deflection of the midpoint chord. and Indicates the distance between adjacent points. This indicates a minimum length. The chord deflection percentage is a dimensionless value.

[0062] The chord deflection rate is generated based on the chord deflection of three adjacent midpoints. The chord deflection rate represents the second-order change ratio of the chord deflection of three adjacent midpoints, used to describe whether a local bend changes abruptly within a continuous window. This calculation requires a five-point window because the current midpoint chord deflection, the previous midpoint chord deflection, and the next midpoint chord deflection each correspond to a combination of three adjacent points.

[0063]

[0064] in, Indicates the rate of change of the chord deflection. , and Both indicate the midpoint of the chord deflection. This indicates a zero-valued length. The rate of change of the chord deflection is a dimensionless value; the larger the rate of change, the more pronounced the abrupt changes in adjacent local geometric results.

[0065] The unidirectionality of the preceding and following segments is positively correlated with the chord deviation stability value, while the chord deviation ratio and chord deviation change rate are negatively correlated with the chord deviation stability value. Higher unidirectionality between the preceding and following segments indicates more continuous local directionality of the radar track; lower chord deviation ratio indicates a smaller relative local path scale of the midpoint chord deviation; and lower chord deviation change rate indicates a smoother bend within adjacent windows. Therefore, this implementation converts the unidirectionality between the preceding and following segments to a non-negative range and suppresses the chord deviation stability value through the chord deviation ratio and chord deviation change rate.

[0066]

[0067] in, Indicates the deflection stability value of the string. Indicates the same directionality between the preceding and following segments. Indicates the chord deflection ratio. This represents the rate of change of chord deflection. The stable value of chord deflection is dimensionless. When the radar track is locally flat and the changes are continuous, the directional similarity between the preceding and following segments is close. When the chord deviation ratio and chord deviation rate are small, the chord deviation stability value increases; when the radar track oscillates significantly in some areas, the forward and backward directions are inconsistent, or the bending changes suddenly increase, the chord deviation stability value decreases.

[0068] Both the chord deflection rate of change and the chord deflection stability value are bound to the current track point within a continuous five-point window. For a radar track point sequence, track points whose beginning and end cannot form a complete continuous five-point window only have their track position points saved; the chord deflection rate of change and chord deflection stability value are not output. Each time the continuous five-point window moves forward by one track point, only one chord deflection stability value is output for the new current track point.

[0069] When five consecutive points are insufficient, the chord deflection stability value generation unit 420 only saves the original track position points and does not output the chord deflection stability value. When the denominator involves the distance between adjacent points, the distance across points, or the sum of chord deflections at the midpoint, the chord deflection stability value generation unit 420 uses either a length zero-prevention variable or an area zero-prevention variable in the calculation. The length zero-prevention variable is set to a positive number less than the minimum effective distance between adjacent points based on the position resolution of the unified coordinate system, and the area zero-prevention variable is set based on the square of the length zero-prevention variable. The length and area zero-prevention variables are only used to prevent division by zero and do not change the trend of the chord deflection stability value changing with the unidirectionality of the preceding and following segments, the proportion of chord deflection, and the rate of change of chord deflection.

[0070] Example 6: The spatiotemporal correlation fusion discrimination unit 430 receives the string-biased stability value output by the string-biased stability value generation unit 420, and determines the radar UAV support degree based on the string-biased stability value. The radar UAV support degree represents the degree of support of the radar track for the UAV category, and its source is the overall level of the string-biased stability value within the current continuous discrimination window and the changes in adjacent string-biased stability values.

[0071] In one implementation, the spatiotemporal correlation fusion discrimination unit 430 selects a continuous discrimination window containing multiple string-biased stable values. The continuous discrimination window can slide along the sampling order of the radar track point sequence, and the number of window points is... The window number represents the number of chord-biased stable values ​​participating in the average calculation within the current radar discrimination window. The window number is an integer greater than or equal to 2, and is determined based on the radar track length, target velocity, and radar sampling period of 100. If the number of effective chord-biased stable values ​​is less than the window number, or the window number is less than 2, then the stable fluctuation value, window normalization value, and radar UAV support will not be output at the current fusion reference time, and the radar category will be marked as unknown.

[0072] A stable average value is generated based on the chord-biased stable values ​​within the continuous discrimination window, and a stable fluctuation value is generated based on the changes in adjacent chord-biased stable values ​​within the continuous discrimination window. The stable average value represents the average result of the chord-biased stable values ​​within the current continuous discrimination window, and the stable fluctuation value represents the average change in adjacent chord-biased stable values ​​within the current continuous discrimination window.

[0073]

[0074] in, Indicates the stable average value. Indicates the number of window points. Indicates the deflection stability value of the string. This indicates the end position of the current continuous discrimination window. The stable average value is a dimensionless value.

[0075]

[0076] in, Indicates the stable fluctuation value. Indicates the number of window points. and Both represent the deflection stability value of the chord. This indicates the end position of the current continuous discrimination window. Stable fluctuation values ​​are dimensionless.

[0077] A window normalization value is generated based on the position of the stable average value relative to the lower and upper stability limits. The lower stability limit represents the lower limit of the stable average value for classifying a radar track as having low support, and the upper stability limit represents the upper limit of the stable average value for classifying a radar track as having high support. The lower and upper stability limits can be written into the data processing computer 400 after statistical analysis of historical samples, or they can be configured during the deployment of the low-altitude surveillance system based on the sampling period, angle measurement accuracy, and target motion samples within the surveillance area of ​​the radar equipment 100.

[0078]

[0079] in, Indicates the window normalization value. Indicates the stable average value. Indicates the lower limit of stability. Indicates the upper limit of stability. This indicates a dimensionless, zero-valued quantity. The window's normalized value is a dimensionless numerical value.

[0080] The support score for the radar-UAV is generated based on the window normalization value and the stability fluctuation value. The window normalization value reflects the position of the stable average value between the lower and upper stability limits, while the stability fluctuation value reflects whether there are significant jumps in the sine-biased stability value within the continuous discrimination window. The higher the stable average value and the lower the stability fluctuation value, the higher the support score for the radar-UAV; conversely, the lower the stable average value or the higher the stability fluctuation value, the lower the support score for the radar-UAV.

[0081]

[0082] in, Indicates the level of support for radar-based drones. Indicates the window normalization value. This represents the stable fluctuation value. The radar UAV support is a dimensionless value, its value is restricted to a non-negative range, and it decreases as the stable fluctuation value increases.

[0083] After generating the radar-UAV support score, when the radar-UAV support score reaches the UAV discrimination threshold, the radar category is marked as a UAV candidate. When the radar-UAV support score is below the bird discrimination threshold and the stable fluctuation value exceeds the bird fluctuation threshold, the radar category is marked as a bird candidate. In other cases, the radar category is marked as unknown. The UAV discrimination threshold, bird discrimination threshold, and bird fluctuation threshold can be written into the data processing computer 400 based on historical sample statistics, the measurement accuracy of radar equipment 100, and the typical target motion characteristics of the low-altitude surveillance area.

[0084] The lower and upper stability limits are determined using historical sample statistics. First, low-support and high-support radar track samples, which have been manually verified or confirmed by multiple devices, are collected. Then, for each sample, the sine-biased stable value, stable average value, and stable fluctuation value are calculated according to this implementation method. The 95th percentile of the stable average value in the low-support radar track samples is then used as the lower stability limit, and the 5th percentile of the stable average value in the high-support radar track samples is used as the upper stability limit. Both the lower and upper stability limits are dimensionless values, and the lower stability limit is less than the upper stability limit. If the calibration results do not satisfy the condition that the lower stability limit is less than the upper stability limit, the sample category is re-verified and samples are re-collected.

[0085] The calibration of the UAV discrimination threshold, bird discrimination threshold, and bird fluctuation threshold adopts a verification sample method. The data processing computer 400 statistically analyzes the radar UAV support and stable fluctuation values ​​of UAV samples, bird samples, and unknown samples in historical samples. The 5th percentile of the radar UAV support for UAV samples is used as the UAV discrimination threshold, the 95th percentile of the radar UAV support for bird samples is used as the bird discrimination threshold, and the 5th percentile of the stable fluctuation value for bird samples is used as the bird fluctuation threshold. After calibration, the fusion category output is checked using verification samples that were not included in the calibration. The misclassification ratio is calculated by dividing the number of verification samples with incorrect output categories by the total number of verification samples. If the misclassification ratio exceeds the calibration misclassification ratio upper limit, samples are re-collected and recalibrated. The calibration misclassification ratio upper limit is read from the low-altitude surveillance system alarm strategy file; if not configured, it is set to 0.05.

[0086] Radar category labels and radar UAV support are stored simultaneously. Radar category labels are used to indicate the target category supported by radar device 100 in subsequent category support calculations, while radar UAV support is used to adjust radar voting weights. By retaining both results simultaneously, radar device 100 not only provides UAV candidates, bird candidates, or unknowns, but also passes the local geometric stability of radar tracks to the fusion category generation process.

[0087] Example 7: The spatiotemporal correlation fusion discrimination unit 430 determines the fusion reference time as the multi-source time alignment benchmark. The fusion reference time can be a predetermined time in a fixed time slice or the sampling time corresponding to the radar track. The predetermined time in the fixed time slice is suitable for scenarios where the data processing computer 400 generates a unified target track with a fixed output period; the sampling time corresponding to the radar track is suitable for scenarios where the radar device 100 is the main time source.

[0088] At each fusion reference time, the spatiotemporal correlation fusion discrimination unit 430 searches for the most recent record in the electronic reconnaissance data and photoelectric data whose time difference is no greater than the time tolerance. The time difference is the difference between the sampling time of the electronic reconnaissance data or photoelectric data and the fusion reference time. The time tolerance is the maximum time difference that allows records from different devices to be included in the same time slice. The most recent record whose time difference is no greater than the time tolerance is included in subsequent spatial correlation, while records whose time difference is greater than the time tolerance are not included in the same target candidate set at the current fusion reference time.

[0089] When electronic reconnaissance data or electro-optical data is not directly recorded at the fusion reference time but exists in valid records before and after it, the spatiotemporal correlation fusion discrimination unit 430 performs linear interpolation. Linear interpolation is only used for position, line-of-sight angle, electronic reconnaissance azimuth angle, or other continuous numerical fields; it does not interpolate electronic reconnaissance category markers or electro-optical category markers. For category fields, if the category markers of two valid records before and after the fusion reference time are consistent, that category marker is used; if the category markers of two valid records are inconsistent, the category marker for that device at the current fusion reference time is set to unknown.

[0090]

[0091] in, Indicates the alignment ratio. Indicates the fusion reference time. Indicates the sampling time of the front side. This indicates the sampling time of the next sample. The alignment ratio is a dimensionless value, and both the sampling time of the previous and next samples are expressed in milliseconds.

[0092]

[0093] in, This represents the value to be aligned corresponding to the fusion reference time. This represents the value to be aligned corresponding to the previous sampling time. This represents the value to be aligned corresponding to the subsequent sampling time. This indicates the alignment ratio. The values ​​to be aligned can be position components, line-of-sight angles, or electronic reconnaissance azimuth angles; the dimensions of the values ​​to be aligned are determined by the field itself.

[0094] When only one side of the electronic reconnaissance data or electro-optical data exists as a valid record, the spatiotemporal correlation fusion discrimination unit 430 performs short-term extrapolation based on the target velocity determined from adjacent valid records. The extrapolation result is stopped once the extrapolation duration exceeds the limit. Short-term extrapolation is used for locatable position fields and is not used for extrapolation of electronic reconnaissance category markers or electro-optical category markers. The one-sided sampling time used for extrapolation can be before or after the fusion reference time; when the one-sided sampling time used for extrapolation is before the fusion reference time, the extrapolation time difference is positive; when the one-sided sampling time used for extrapolation is after the fusion reference time, the extrapolation time difference is negative.

[0095]

[0096] in, Indicates the time difference between adjacent valid records. Indicates the sampling time of the front side. Indicates the sampling time of the next sample. The time difference between adjacent valid records is measured in seconds.

[0097]

[0098] in, This indicates the target velocity determined by adjacent valid records. This indicates the track position point corresponding to the previous sampling time. This indicates the track position point corresponding to the subsequent sampling time. This represents the time difference between adjacent valid records. The target velocity determined by adjacent valid records is measured in meters per second.

[0099]

[0100] in, Indicates the extrapolated time difference. Indicates the fusion reference time. This indicates the single-sided sampling time used for extrapolation. The extrapolation time difference is measured in seconds and can be positive, zero, or negative.

[0101]

[0102] in, Indicates the extrapolation duration. This represents the extrapolated time difference. The extrapolated duration is a non-negative number, measured in seconds.

[0103]

[0104] in, Indicates the position of the extrapolated track. This represents the track position point corresponding to the one-sided sampling time used for extrapolation. This indicates the target velocity determined by adjacent valid records. This represents the extrapolated time difference. The units of the extrapolated track position points are meters. When the one-sided sampling time used for extrapolation is earlier than the fusion reference time, The value is positive and extrapolated along the target velocity direction determined by adjacent valid records; when the single-sided sampling time used for extrapolation is after the fusion reference time, The value is negative and extrapolated in the opposite direction to the target velocity determined by the adjacent valid records. The extrapolation duration limit is determined based on the equipment sampling period, the possible target velocity, and the prediction error allowed by the low-altitude surveillance system; when the extrapolation duration exceeds the extrapolation duration limit, the extrapolated track position point will not enter the same target candidate set.

[0105] For each record that has completed multi-source time alignment, the spatiotemporal correlation fusion discrimination unit 430 saves the fusion reference time, the original sampling time, and the time difference. The fusion reference time is used to unify the current processing cycle, the original sampling time is used to trace the device output time, and the time difference is used to calculate the deviation of candidate combinations in the subsequent formation of candidate sets for the same target.

[0106] Example 8: The spatiotemporal correlation fusion discrimination unit 430 forms a candidate set of the same target based on spatial intervals and correlation thresholds. The candidate set of the same target represents a combination of multi-source records that may belong to the same physical target at the same fusion reference time. Each candidate set of the same target contains at least one radar track record and may contain one electronic reconnaissance marker and one electro-optical marker.

[0107] Spatial intervals are calculated for each radar track, electronic reconnaissance azimuth line, and electro-optical target position within the same fusion reference time. For the positions of locatable equipment targets, the spatial interval can be the Euclidean distance in a unified coordinate system. For the electronic reconnaissance azimuth line, the spatial interval can be the lateral distance from the corresponding position of the radar track to the electronic reconnaissance azimuth line. For the optical television axis azimuth, the spatial interval can be determined by combining the azimuth difference between the relative azimuth of the radar track and the optical television axis azimuth.

[0108]

[0109] in, Indicates spatial interval, and Indicates the eastward position. and Indicates the northward position. and Indicates vertical position. The unit of measurement for spatial intervals is meters.

[0110]

[0111] in, Indicates spatial interval, This indicates the association threshold. The association threshold represents the maximum spatial interval at which two tracks are allowed to merge into the same target. The association threshold can be written into the data processing computer 400 based on equipment measurement errors, the target's possible speed, and timing tolerance.

[0112] Data whose spatial interval is no greater than the correlation threshold, whose altitude difference between the radar track and the location of the locatable equipment target does not exceed the altitude difference limit, and whose azimuth difference between the relative azimuth of the radar track and the azimuth of the electronic reconnaissance azimuth line or optical television axis does not exceed the azimuth difference limit, are grouped into the same target candidate set. The altitude difference limit and azimuth difference limit are determined based on the equipment installation location, orientation calibration value, angle measurement accuracy, and the target altitude range of the low-altitude surveillance area.

[0113] When a device record falls into multiple candidate sets of the same target simultaneously, the spatiotemporal correlation fusion discrimination unit 430 generates a candidate combination deviation degree characterized by the deviation degree of the spatial interval relative to the correlation threshold, the time difference, and the heading difference, and retains the candidate combination with the smallest deviation degree. The deviation degree of the spatial interval relative to the correlation threshold is used to represent the spatial proximity, the time difference is used to represent the temporal proximity, and the heading difference is used to represent the proximity of the target's motion direction.

[0114]

[0115] in, Indicates the spatial deviation. Indicates spatial interval, Indicates the associated threshold, This indicates a minimum length. Spatial deviation is a dimensionless value.

[0116]

[0117] in, Indicates the time deviation. Indicates time difference, Indicates time tolerance. This indicates a time deviation from zero. The time deviation is a dimensionless value, while the time deviation from zero is measured in milliseconds.

[0118]

[0119] in, Indicates the deviation from the heading. Indicates the difference in heading. Indicates the azimuth difference limit. This indicates the angle zero-prevention quantity. The heading deviation is a dimensionless value, while the heading difference, bearing difference limit, and angle zero-prevention quantity are all in radians.

[0120]

[0121] in, Indicates the degree of deviation of candidate combinations. Indicates the spatial deviation. Indicates the time deviation. Indicates the deviation from the heading. Indicates spatial deviation weights, Indicates the time deviation weight. This represents the heading deviation weight. The spatial deviation weight, temporal deviation weight, and heading deviation weight are all dimensionless values, and their sum is 1; when not configured individually, all three take the same value.

[0122] The heading difference can be determined by the positions of adjacent track points. When the equipment record lacks adjacent track positions that can be used to calculate the heading difference, the heading deviation is set to 1, so that the combination does not receive extra priority due to the lack of heading difference. If the candidate combinations of two combinations have the same degree of deviation, the combination with the smaller time difference is preferred; if the time differences are still the same, the combination with the smaller spatial interval is preferred; if the spatial intervals are still the same, the combination corresponding to the equipment record with the earlier reception time is preferred.

[0123] For the same set of candidate targets with one-to-many or many-to-one relationships, the spatiotemporal correlation fusion discrimination unit 430 uses nearest neighbor order processing and checks that the same device record cannot be assigned to two unified targets at the same fusion reference time. Combinations that are not retained are marked as correlation conflicts, and the fusion reference time where the correlation conflict occurred, the involved devices, and the degree of deviation of the candidate combination are saved. This processing is used to maintain the stability and traceability of the unified target track.

[0124] Example 9: After forming the same target candidate set, the spatiotemporal correlation fusion discrimination unit 430 reads the radar category marker, electronic reconnaissance category marker, and electro-optical category marker from the same target candidate set, and generates category support based on their respective current voting weights. Target categories include UAVs, birds, and unknowns. The radar category marker is determined by the radar UAV support and stable fluctuation value, the electronic reconnaissance category marker comes from the output of the electronic reconnaissance equipment 200 or its post-processing result, and the electro-optical category marker comes from the output of the electro-optical equipment 300 or its post-processing result.

[0125] When adjusting radar voting weights using radar drone support, the radar voting weights are jointly determined by radar device credibility and radar drone support. The radar voting weights increase as radar drone support increases and decrease as radar drone support decreases. Radar device credibility represents the basic credibility of radar device 100 in the fusion voting; it is the device credibility when the device source is radar device 100.

[0126] Equipment reliability is determined based on the proportion of historical valid detections, the proportion of false associations, the proportion of correct category label verifications, and the proportion of equipment online. The statistical sample comes from raw equipment data stored on the data processing computer 400, the same target candidate set, manual verification results, multi-device consistency confirmation results, and equipment operation logs. The statistical period uses the most recent complete calibration period; the calibration period is read from the low-altitude monitoring system configuration file, or, if not configured, the most recent 30 days of operation records. Equipment reliability is recalculated when equipment maintenance, changes in equipment installation location, changes in orientation calibration values, changes in the equipment data dictionary, or when the misjudgment rate exceeds the upper limit of the calibration misjudgment rate.

[0127]

[0128] in, This indicates the percentage of historically valid tests conducted by the equipment. Indicates the first The number of records of a type of device that were mapped as valid quality indicators within the statistical period. Indicates the first The total number of records for this type of device within the statistical period. This indicates dimensionless protection against zero-value defects. The historical effective detection rate of the equipment is a dimensionless value.

[0129]

[0130] in, Indicates the proportion of false associations. Indicates the first The number of records of a device type that were determined to be incorrectly associated after manual review or consensus confirmation among multiple devices within the statistical period. Indicates the first The number of records of the same type of device entering the same target candidate set within the statistical period. This indicates a dimensionless, zero-valued quantity. The erroneous correlation ratio is a dimensionless numerical value.

[0131]

[0132] in, This indicates the percentage of categories that have been correctly verified. Indicates the first The number of records where the category label of a type of equipment matches the category manually verified during the statistical period. Indicates the first The number of records for which category verification has been completed for each type of equipment within the statistical period. This indicates a dimensionless, zero-valued quantity. The category markers and verification of correct proportions are dimensionless numerical values.

[0133]

[0134] in, Indicates the percentage of devices online. Indicates the first The online duration of this type of device within the statistical period. Indicates the duration of the statistical period. This indicates the time limit for preventing zero-value transactions. Online duration, statistical period duration, and time limit for preventing zero-value transactions are all expressed in milliseconds. The device online percentage is a dimensionless value.

[0135]

[0136] in, Indicates the reliability of the equipment. This indicates the lower limit of device reliability. Indicates the effective detection ratio weight. Indicates the proportion and weight of misassociations. Indicates the category review ratio weight. This represents the online proportion weight. The device credibility lower limit, effective detection proportion weight, false association proportion weight, category review proportion weight, and online proportion weight are all dimensionless values, and the sum of the four weights is 1; when not configured separately, the four weights take the same value, and the device credibility lower limit is 0.1.

[0137]

[0138] in, Indicates the reliability of radar equipment. Indicates the reliability of the equipment. This indicates that the equipment originates from radar equipment 100. The reliability of the radar equipment, electronic reconnaissance equipment 200, and optoelectronic equipment 300 uses the same normalized interval, with a value range of 0 to 1.

[0139]

[0140] in, Indicates the radar voting weight. Indicates the reliability of radar equipment. This indicates the support level for radar-based drones. The radar voting weight is a dimensionless value used to determine the degree of influence of the radar category label on the fusion category at the current fusion reference time.

[0141] The current voting weights of electronic reconnaissance equipment 200 and optoelectronic equipment 300 are determined based on equipment credibility, spatial interval, time difference, and equipment quality indicator. The current voting weight is the voting weight corresponding to each equipment category label; equipment credibility represents the basic credibility level of the corresponding equipment in the fusion voting. Spatial interval reflects the spatial proximity of the equipment's record to the radar track. Time difference reflects the temporal proximity of the equipment's record to the fusion reference time. Equipment quality indicator reflects the detection or tracking status of the equipment in the current record. The above current voting weights do not change the calculation method of the radar voting weights.

[0142]

[0143] in, Indicates the space reduction factor. Indicates the first The spatial interval between the device's record and the radar track. Indicates the associated threshold, This indicates a length that is zero. The space reduction factor is a dimensionless value.

[0144]

[0145] in, Indicates the time reduction factor. Indicates the first The time difference between the device's record and the fusion reference time. Indicates time tolerance. This indicates a time reduction factor that is zero. The time reduction factor is a dimensionless value.

[0146]

[0147] in, Indicates the current voting weight. Indicates the reliability of the equipment. Indicates the space reduction factor. Indicates the time reduction factor. This represents the quality reduction factor. The current voting weight is a dimensionless value. The quality reduction factor is obtained by mapping the equipment quality flag; the quality reduction factor is 1 when the equipment quality flag is valid, 0.5 when the equipment quality flag is downgraded, and 0 when the equipment quality flag is invalid or the equipment is missing data. The equipment quality flags of radar equipment 100, electronic reconnaissance equipment 200, and optoelectronic equipment 300 are all generated according to the aforementioned three-level mapping rule, so that different original quality fields enter the same quality reduction factor value range.

[0148] Based on radar category labels, electronic reconnaissance category labels, and electro-optical category labels and their current voting weights, support scores for UAVs, birds, and unknown categories are generated respectively. The category indication value indicates whether a device supports the target category. When the device category label equals the target category, the category indication value for that target category is 1; when the device category label does not equal the target category, the category indication value for that target category is 0. If the device category label is unknown, the category indication value for the unknown target category is 1, while the category indication values ​​for UAVs and birds are 0.

[0149]

[0150] in, Indicates the category indicator value. Indicates the first Equipment category label for class devices, Indicates the target category. The category indicator value is a dimensionless numerical value.

[0151]

[0152] in, Indicates category support. Indicates the target category, Indicates participation in the equipment set. Indicates the current voting weight. Indicates the category indicator value. This indicates a dimensionless zero-valued quantity. The category support is a dimensionless numerical value.

[0153] The spatiotemporal correlation fusion discrimination unit 430 selects the category corresponding to the highest category support as the fusion category. When the highest category support is lower than the minimum output threshold for category support, or when the difference in category support between the highest and second-highest category support is less than the category support difference confirmation threshold, the fusion category is set to unknown. The minimum output threshold for category support is used to avoid outputting a definite category when the overall support of multi-source data is insufficient, and the category support difference confirmation threshold is used to avoid prematurely determining the fusion category when the support between drones and birds is close.

[0154] Example 10: The spatiotemporal correlation fusion discrimination unit 430 compares the current fusion category with the fusion category of the same unified target at a previous fusion reference time, and confirms the current fusion category based on the continuous consistency result. If the current fusion category is consistent with the fusion category at a previous fusion reference time, the continuity count is increased. If the current fusion category changes, it first enters the pending confirmation state. Only when the new category obtains the highest category support at multiple consecutive fusion reference times is the official category of the unified target updated; otherwise, the previous official category is retained.

[0155] For newly emerging unified targets, an "unknown" output can be generated until the minimum number of consecutive observations is reached. The minimum number of consecutive observations is determined based on the output cycle of the low-altitude surveillance system and false alarm control requirements. For unified targets that reappear after a short period of loss and whose position and velocity meet the conditions for continuation of the original track, the data processing computer 400 can inherit the original unified target's category history and consecutive count. This processing is used to improve the consistency of the fused category for the same low-altitude target within consecutive time slices.

[0156] When generating a unified target track, the data processing computer 400 assigns a unified target number to the same set of target candidates through multi-source time alignment, spatial correlation, and category fusion, and records the associated radar track number, electronic reconnaissance marker number, and electro-optical target number. The unified target track includes the unified target number, fusion reference time, three-dimensional position, three-dimensional velocity, fusion category, category support, list of associated equipment, and data quality status.

[0157] The 3D position and 3D velocity in the unified target track are determined according to the number of valid devices. If only one device in the same target candidate set can provide a valid 3D position, the unified target track adopts the 3D position of that device. If two or more devices in the same target candidate set can provide a valid 3D position, the unified target track adopts a weighted average of the current voting weights. If no device can provide a valid 3D position, the 3D position in the unified target track is not updated at the current fusion reference time, and the data quality status is marked as position missing.

[0158]

[0159] in, Indicates a unified target location. This represents the set of participating devices that can provide valid three-dimensional positions. Indicates the current voting weight. Indicates the first The three-dimensional position provided by the device This indicates a dimensionless, zero-valued quantity. The unit of measurement for the target location is the meter.

[0160]

[0161] in, Indicates a unified target speed. This represents the set of participating devices capable of providing state-valid three-dimensional velocities. Indicates the current voting weight. Indicates the first The three-dimensional velocity provided by the device This indicates a dimensionless, zero-valued quantity. The dimension of the target velocity is meters per second.

[0162] The unified target track preserves the correlation between radar data, electronic reconnaissance data, and optoelectronic data, so that the monitoring display terminal 500 displays only one unified target marker in the map or 3D scene, while retaining the original markers of each device in the details window for operators to verify.

[0163] The data processing computer 400 outputs the unified target track, fusion category, and category support to the monitoring and display terminal 500, and uses the above results as the final display and storage results of the current fusion reference time.

[0164] Example 11: When electronic reconnaissance equipment 200 or optoelectronic equipment 300 is temporarily missing, the data processing computer 400 continues to use radar tracks to perform target maintenance and removes the missing equipment from the set of participating equipment at this fusion reference time. The missing equipment does not generate current voting weights, nor does it participate in the current category support calculation. If the missing equipment resumes output at a subsequent fusion reference time, the newly received data will be re-processed with field unification, spatial coordinate unification, multi-source time alignment, and spatial correlation.

[0165] When radar track detection is unavailable, the data processing computer 400 maintains the unified target briefly only if other devices can locate it and the continuity condition is met. During this brief maintenance period, the data processing computer 400 does not generate new string deflection stability values ​​or update the radar UAV support. If a valid radar track cannot be obtained after the permitted brief maintenance time, the unified target enters a state of loss of contact or termination.

[0166] When a unified target fails to receive any valid equipment update within the specified termination time, the data processing computer 400 changes the target status to terminated, stops further extrapolation, and saves the last location, last category, and termination reason. The termination time is configured based on the low-altitude surveillance system output cycle, equipment sampling cycle, and the target's possible speed. After equipment recovery, newly received data does not directly use the equipment records from before the loss of connection to prevent expired data from participating in fusion again.

[0167] The data processing computer 400 writes raw equipment data, standardized format data, multi-source time alignment results, spatial correlation results, candidate sets for the same target, radar category markers, electronic reconnaissance category markers, electro-optical category markers, fused categories, and the final unified target track into the database according to a unified target number. It establishes an event log for each category change, correlation conflict, equipment disconnection, and track split. The event log includes the time of occurrence, the equipment involved, the pre-processing state, and the post-processing state.

[0168] As an optional implementation, radar device 100 continuously outputs range, azimuth, and elevation angles, while electronic reconnaissance device 200 and optoelectronic device 300 both output valid records within the time tolerance. The radar track point sequence, after cleaning and sorting, forms a sufficient number of consecutive five-point windows. The chord deviation stability value generation unit 420 generates a small midpoint chord deviation, a low chord deviation percentage, a low chord deviation rate of change, and high unidirectionality between the preceding and following segments, resulting in a high chord deviation stability value. The stable average value within the continuous discrimination window is high, and the stable fluctuation value is low. The spatiotemporal correlation fusion discrimination unit 430 generates a high radar UAV support score and allows the radar voting weight to play a significant role in the generation of the fusion category.

[0169] As an optional implementation, low-altitude targets may exhibit localized oscillations or reversals within a short period. Using only a single fitting residual or a single turning angle can lead to numerical scale instability due to changes in target distance and the distance between adjacent points. This implementation describes the magnitude of local bending using midpoint chordal deviation, normalizes the midpoint chordal deviation locally using its proportion, describes the second-order change of the chordal deviation at three adjacent midpoints using the rate of change of chordal deviation, and limits turning-around or reversal scenarios by using the unidirectionality of the preceding and following segments. In this implementation, the stability value of the chordal deviation decreases as the proportion and rate of change of chordal deviation increase, leading to a decrease in radar / UAV support and radar voting weight. Electronic reconnaissance and electro-optical category markers play a more significant role in the fusion category generation.

[0170] As an optional implementation, electronic reconnaissance equipment 200 may temporarily be undetected at a certain fusion reference moment. Data processing computer 400 removes electronic reconnaissance equipment 200 from the current set of participating equipment and continues to generate category support using the valid records of radar equipment 100 and optoelectronic equipment 300. If optoelectronic equipment 300 is also temporarily undetected but radar tracks remain valid, data processing computer 400 continues to use radar tracks to perform target maintenance and outputs the current results according to radar UAV support and radar category labeling. If radar tracks are undetected, data processing computer 400 maintains a unified target only briefly when other equipment can locate it and the continuity condition is met.

[0171] As an optional implementation, if the five consecutive point windows are insufficient or the distance between points, the distance between adjacent points, and the sum of the midpoint chordal deviations involved in the denominator are too small, the chordal deviation stable value generation unit 420 will not force the output of unreliable results. When the five consecutive point windows are insufficient, only the original track position points are saved; when the denominator is too small, anti-zero decimals for length, area, time, angle, or dimensionless quantities are introduced according to the formula dimensions. If the highest class support is lower than the minimum output threshold for class support, or the difference in class support between the highest and second-highest class support is less than the class support difference confirmation threshold, the fused class is set to unknown. This processing avoids boundary data directly causing the determination of the class output.

[0172] As an optional implementation, the minimum output threshold for category support and the confirmation threshold for category support difference are calibrated based on fused category validation samples. The data processing computer 400 calculates the category support for the drone samples, bird samples, and unknown samples respectively. The 95th percentile of the highest category support among the unknown samples is used as the minimum output threshold for category support, and the 5th percentile of the category support difference between the highest and second-highest category support among the correctly identified category samples is used as the confirmation threshold for category support difference. Both the minimum output threshold for category support and the confirmation threshold for category support difference are dimensionless values, ranging from 0 to 1.

[0173] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A spatiotemporal fusion discrimination method for low-altitude target radar tracks, applied to a data processing computer, wherein the data processing computer receives radar data, electronic reconnaissance data, and optoelectronic data, and performs field unification and spatial coordinate unification on the various source data, characterized in that... The method includes: The radar track position points in the radar track point sequence are generated based on the range, azimuth and elevation angles output by the radar equipment. The radar track point sequence is cleaned and sorted. A continuous five-point window is established in the radar track point sequence, and a stable chord deviation value is generated based on the midpoint chord deviation, chord deviation rate of change, and unidirectionality of the preceding and following segments in the continuous five-point window. A fusion reference time is determined as the multi-source time alignment benchmark, and electronic reconnaissance data and photoelectric data are time aligned from multiple sources. The same target candidate set is formed according to the spatial interval and correlation threshold. The support of radar UAVs and radar category labels are determined based on the string deflection stability value. The radar voting weights are adjusted using the support of radar UAVs, and the fused category is obtained by combining the radar category label, electronic reconnaissance category label and optoelectronic category label. The current fusion category is compared with the fusion category of the same unified target at the previous fusion reference time. Based on the continuous and consistent results, the current fusion category is confirmed, a unified target track is generated, and the fusion category and category support are output.

2. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, When receiving radar data, electronic reconnaissance data, and optoelectronic data, the data processing computer writes the device number and reception time to each piece of data and distinguishes radar data, electronic reconnaissance data, and optoelectronic data according to the device source. Among them, the device number is used to distinguish different device sources in subsequent cleaning and sorting, and the reception time is used to determine the retention order of duplicate records when the quality indicators are the same.

3. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, Unifying fields and spatial coordinates includes unifying distance units to meters, angle units to radians, and time units to milliseconds; for device data that has already been output in rectangular coordinates, it is converted to a unified coordinate system according to the device installation location, orientation calibration value, and coordinate transformation matrix.

4. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, Cleaning and sorting the radar track point sequence includes: Records lacking track number, sampling time, distance measurement value, azimuth angle, or pitch angle are marked as unusable for association; records exceeding the equipment's measurement range or angle range are marked as abnormal records. For records containing a speed field, records exceeding the speed range are marked as abnormal records; and abnormal records are excluded from the formation of the same target candidate set. Sort by device number, track number, and sampling time; For duplicate records of the same device, the same track number, and the same sampling time, the record with the higher quality flag is retained; if the quality flags are the same, the record with the earlier reception time is retained.

5. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, The string deflection stability value is generated based on a five-point window, including: In a five-point window, determine the current waypoint, its two preceding waypoints, and its two following waypoints; Generate entry and exit line segment vectors based on adjacent track position points; Generate the cross-point distance, adjacent point spacing, midpoint chord deflection, and front and rear segment homing based on the entry and exit line segment vectors; The chord deviation ratio is generated based on the distance between adjacent points and the chord deviation of the midpoint. The rate of change of chord deflection is generated based on the chord deflection of three adjacent midpoints; Make the unidirectionality of the front and rear segments positively correlated with the chord deflection stability value, and make the chord deflection ratio and chord deflection change rate negatively correlated with the chord deflection stability value.

6. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, Determining radar UAV support based on string deflection stability values ​​includes: Select a continuous discrimination window that contains multiple string deflection stability values; A stable average value is generated based on the chord deviated stable value within the continuous discrimination window, and a stable fluctuation value is generated based on the changes in adjacent chord deviated stable values ​​within the continuous discrimination window. Window normalization values ​​are generated based on the position of the stable average relative to the lower and upper limits of stability. Radar UAV support is generated based on the window normalization value and the stable fluctuation value.

7. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 6, characterized in that, After generating the radar drone support score, when the radar drone support score reaches the drone discrimination threshold, the radar category is marked as a drone candidate; when the radar drone support score is lower than the bird discrimination threshold and the stable fluctuation value exceeds the bird fluctuation threshold, the radar category is marked as a bird candidate. In all other cases, the radar category will be marked as unknown.

8. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, Multi-source time alignment includes: The predetermined time in the fixed time slice or the sampling time corresponding to the radar track is used as the fusion reference time; Search for the most recent record in electronic reconnaissance data and photoelectric data whose time difference is no greater than the time tolerance; Linear interpolation is performed when electronic reconnaissance data or photoelectric data is not directly recorded at the fusion reference time but has valid records before and after it. When electronic reconnaissance data or photoelectric data has only one valid record, the target velocity determined by the adjacent valid record is extrapolated for a short time, and the extrapolation result is stopped after the extrapolation time exceeds the extrapolation time limit.

9. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, The same set of candidate targets is formed based on spatial intervals and correlation thresholds, including: The spatial intervals of radar tracks, electronic reconnaissance azimuth lines, and electro-optical target positions within the same fusion reference time are calculated one by one. Data whose spatial interval is not greater than the correlation threshold, whose altitude difference between the radar track and the target position of the locatable device does not exceed the altitude difference limit, and whose azimuth difference between the relative azimuth of the radar track and the azimuth of the electronic reconnaissance azimuth line or the optical television axis does not exceed the azimuth difference limit are grouped into the same target candidate set. When a device record falls into multiple candidate sets of the same target at the same time, a candidate combination deviation degree is generated, which is characterized by the deviation degree of the spatial interval relative to the associated threshold, the time difference, and the heading difference, and the combination with the smallest candidate combination deviation degree is retained.

10. The spatiotemporal fusion discrimination method for low-altitude target radar tracks according to claim 1, characterized in that, Adjusting radar voting weights using radar drone support includes: The radar voting weight is determined by the credibility of radar equipment and the support of radar drones, so that the radar voting weight increases as the support of radar drones increases and decreases as the support of radar drones decreases. Based on radar category labels, electronic reconnaissance category labels, and optoelectronic category labels and their corresponding voting weights, the category support for UAVs, birds, and unknown categories is generated respectively, and the category corresponding to the highest category support is selected as the fusion category; When the highest category support is lower than the minimum output threshold for category support, or when the difference in category support between the highest and second-highest category support is less than the confirmation threshold for category support difference, the fused category is set to unknown.

Images