Unmanned aerial vehicle low false alarm identification method based on radar multi-dimensional micro-motion characteristics

Abstract

The invention discloses an unmanned aerial vehicle low false alarm identification method based on radar multi-dimensional micro-motion features, belongs to the technical field of radar target identification, and is used for solving the technical problems that in existing unmanned aerial vehicle low-altitude target identification, the identification performance of a target model is difficult to guarantee, and an extremely low false alarm rate of a target is difficult to realize. The method comprises the following steps: according to a target micro-motion Doppler spectrum, carrying out multi-dimensional feature extraction processing on a target main body signal and a micro-motion signal at a low altitude to obtain a target multi-dimensional micro-motion feature; performing setting processing of multi-level discrimination thresholds on the multi-dimensional micro-motion features of the target, and performing series connection to obtain a series connection type three-level threshold determination mechanism; based on a threshold judgment result of the tandem type three-level threshold judgment mechanism, obtaining a target recognition result under the target track; and through a preset dynamic sliding window, performing time sequence fusion and target identification optimization processing on the single target identification results in the track hierarchy to obtain an optimized target identification result in the same target track.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of radar target recognition, in particular to a low false alarm recognition method for unmanned aerial vehicles based on radar multi-dimensional micro-motion characteristics. BACKGROUND

[0002] Low-altitude target type recognition plays an important supporting role in aviation safety guarantee and low-altitude situation awareness research, and is one of the key tasks of radar low-altitude target detection. However, most of the existing radar recognition methods for unmanned aerial vehicles use signal decomposition or time-frequency transformation methods to extract features, and combine machine learning or deep learning algorithms to realize target classification. Such methods have obvious limitations: on the one hand, the intelligent recognition method based on data has limited generalization ability for target models outside the database, and the recognition performance is difficult to guarantee; on the other hand, such methods usually optimize the high recognition accuracy, but it is difficult to achieve extremely low false alarm rate in actual application, thereby causing problems such as data sensitivity and insufficient false alarm control ability of the system.

[0003] Therefore, it is urgent to develop a low false alarm recognition method for unmanned aerial vehicle targets based on radar multi-dimensional micro-motion characteristics, which can effectively extract radar discrimination features according to the micro-motion characteristics of low-altitude targets, and effectively control the false alarm while achieving high-precision recognition. SUMMARY

[0004] The embodiment of the application provides a low false alarm recognition method for unmanned aerial vehicles based on radar multi-dimensional micro-motion characteristics, which is used to solve the following technical problems: in the existing low-altitude target recognition of unmanned aerial vehicles, the recognition performance of the target model is difficult to guarantee, and it is difficult to achieve an extremely low false alarm rate for the target.

[0005] The embodiment of the application adopts the following technical scheme: On the one hand, the embodiment of the application provides a low false alarm recognition method for unmanned aerial vehicles based on radar multi-dimensional micro-motion characteristics, which comprises: according to the target micro-motion Doppler spectrum, multi-dimensional feature extraction and processing are performed on the multi-dimensional characteristics between the target main body signal and the micro-motion signal under low altitude, to obtain target multi-dimensional micro-motion characteristics; wherein the target multi-dimensional micro-motion characteristics include harmonic interval characteristics, harmonic isolation characteristics and main harmonic ratio characteristics; according to the micro-motion characteristics of low-altitude targets, the target multi-dimensional micro-motion characteristics are subjected to multi-stage threshold setting processing, and a series of three-stage threshold determination mechanisms are obtained in series; based on the threshold determination result of the series of three-stage threshold determination mechanisms, a target recognition result under the target track is obtained; through a pre-set dynamic sliding window, the single target recognition result in the track level is subjected to time sequence fusion and target recognition optimization processing, to obtain an optimized target recognition result in the same target track.

[0006] The embodiments of the present application can realize high accuracy and low false alarm recognition of low-altitude unmanned aerial vehicle targets through low false alarm recognition of the unmanned aerial vehicle based on radar multi-dimensional micro-motion characteristics. The embodiments can effectively adapt to the changes in target motion state and target model, and significantly improve the discrimination ability of unmanned aerial vehicles and non-unmanned aerial vehicle targets. Especially when facing out-of-database targets, the proposed recognition based on target micro-motion physical characteristics does not rely on data-driven mode, thereby effectively reducing the dependence of the recognition method on sample data. At the same time, through the dynamic sliding window optimization strategy, the track level can also enhance the time sequence consistency of the recognition result, and improve the reliability and practicality of the system in complex low-altitude environment.

[0007] In a feasible implementation, before the multi-dimensional characteristics between the target main body signal and the micro-motion signal under low altitude are extracted and processed according to the target micro-motion Doppler spectrum to obtain the target multi-dimensional micro-motion characteristics, the method further includes: performing target detection processing on the distance-Doppler plane of the current track point echo through a low false alarm rate CFAR detector to obtain a first target detection result; performing target micro-motion harmonic signal detection processing on the target in the current track point based on the Doppler domain and through a two-stage CFAR detector to obtain a second target detection result; and performing detection fusion extraction processing on the target with respect to the target main body signal and the micro-motion signal according to the first target detection result and the second target detection result to obtain a target Doppler domain detection result.

[0008] In an implementation, according to the target micro-Doppler spectrum, a multi-dimensional feature extraction process is performed between the target main body signal and the micro-motion signal under low altitude, to obtain a target multi-dimensional micro-motion feature, which specifically includes: according to the target Doppler domain detection result, and based on the target micro-Doppler spectrum, the harmonic with the smallest interval from the main Doppler frequency is selected; and the smallest harmonic and the main Doppler frequency are difference calculated to obtain the fundamental frequency; based on the fundamental frequency, a search process is performed on whether there is a high-order harmonic in the preset absolute frequency error tolerance, to detect and obtain a plurality of high-order harmonic points, and the high-order harmonic points are included in a harmonic set; the harmonic set and the main Doppler are jointly processed to obtain the harmonic interval feature of the target; based on the harmonic interval feature, a pair of adjacent harmonic points of the target are located; the average power of the pair of adjacent harmonic points is calculated and determined as the harmonic power; the average power of adjacent signals in the pair of adjacent harmonic points is calculated and determined as the noise floor power; the harmonic power and the noise floor power are difference calculated to obtain the energy focusing metric of the pair of adjacent harmonic points; all adjacent harmonic pairs are traversed to obtain a set of all power difference values based on different energy focusing metrics; and the set of all power difference values is determined as the harmonic isolation feature of the target; based on the amplitude decibel value of each harmonic in the harmonic interval feature, the main harmonic ratio feature is obtained; the harmonic interval feature, the harmonic isolation feature, and the main harmonic ratio feature are combined into a multi-dimensional feature to obtain the target multi-dimensional micro-motion feature.

[0009] In an implementation, based on the amplitude decibel value of each harmonic in the harmonic interval feature, the main harmonic ratio feature is obtained, which specifically includes: obtaining the amplitude decibel value of each harmonic in the harmonic interval feature; the amplitudes of each harmonic in the harmonic interval feature are numerically compared, and the harmonic with the largest amplitude is determined as the representative harmonic of the micro-Doppler; the target main Doppler and the maximum harmonic amplitude decibel value of the representative harmonic are difference calculated to obtain the main harmonic ratio feature.

[0010] In an embodiment, the target multi-dimensional micro-motion characteristics are subjected to multi-stage threshold setting processing according to the low-altitude target micro-motion characteristics, and a series of three-stage threshold determination mechanisms are obtained in series, specifically including: according to the UAV micro-motion model, determining whether the target Doppler spectrum contains a preset number of harmonics to determine the target category, and based on a first determination rule, a first-stage threshold determination mechanism based on the harmonic interval characteristics is constructed; the first-stage threshold determination mechanism includes: potential UAV and non-UAV; according to the high signal-to-noise ratio and narrow band characteristics of the UAV harmonic components, the threshold of the harmonic signal of the target under high signal-to-noise ratio and narrow bandwidth is determined, and based on a second determination rule, a second-stage threshold determination mechanism based on the harmonic isolation characteristics is constructed; the second-stage threshold determination mechanism includes: potential UAV and non-UAV; according to the radar scattering interface of the UAV body, the power ratio between the target main Doppler component and the strongest harmonic component of the target is calculated to obtain the minimum main harmonic ratio; based on the threshold judgment condition of the minimum main harmonic ratio and a third determination rule, a third-stage threshold determination mechanism based on the main harmonic ratio characteristics is obtained; the third-stage threshold determination mechanism includes: potential UAV and non-UAV; the first-stage threshold determination mechanism, the second-stage threshold determination mechanism, and the third-stage threshold determination mechanism are sequentially subjected to threshold series processing to obtain the series of three-stage threshold determination mechanisms.

[0011] In an embodiment, based on the threshold determination result of the series of three-stage threshold determination mechanisms, a target recognition result under a target track is obtained, specifically including: the target multi-dimensional micro-motion characteristics of the target are sequentially subjected to threshold determination through the series of three-stage threshold determination mechanisms; if potential UAVs are present in each threshold determination mechanism, the threshold determination result is determined as a UAV recognition result; if non-UAVs are present in each threshold determination mechanism, the threshold determination result is determined as a non-UAV determination result; based on the UAV recognition result and the non-UAV determination result, the target recognition result under the target track is obtained.

[0012] In an embodiment, through a preset dynamic sliding window, a single target recognition result in a track level is subjected to time sequence fusion and target recognition optimization processing to obtain an optimized target recognition result in the same target track, specifically including: through the dynamic sliding window, each target recognition result in the same track level is subjected to sliding strategy control; the historical sliding window optimization result in the dynamic sliding window is used as a window reference value; through the window reference value, a recursive recognition update is performed on the single target recognition result in the same target track; and based on the historical sliding window optimization result, a consistency constraint is performed on the current track target recognition result to complete an abnormal fluctuation correction processing of the single target recognition result, and the optimized target recognition result of each track point in the same target track is obtained.

[0013] In an implementable embodiment, according to , a fusion judgment condition is obtained ; wherein, is the dynamic sliding window; is a reference window weight factor; is the historical target recognition result in the historical sliding window optimization result at the moment t; t is the moment corresponding to the track point in the same target track; according to , a fusion decision function is obtained; wherein, the is the historical target recognition result, and ; according to , the optimization target recognition result is obtained; wherein, is the original target recognition result in the track level.

[0014] The application provides a low false alarm recognition method for unmanned aerial vehicles based on radar multi-dimensional micro-motion features, and compared with the prior art, the embodiments of the application have the following beneficial technical effects: 1. The low false alarm recognition for low-altitude unmanned aerial vehicle targets can be realized with high accuracy.

[0015] 2. The method can effectively adapt to the changes in target motion state and target model, and significantly improve the discrimination ability for unmanned aerial vehicle and non-unmanned aerial vehicle targets.

[0016] 3. When facing out-of-database targets, the method is based on target micro-motion physical features for recognition, and does not rely on data-driven mode, thereby effectively reducing the dependence of the recognition method on sample data.

[0017] 4. The dynamic sliding window optimization strategy can also be used to enhance the time sequence consistency of the track level recognition result, and improve the reliability and practicability of the system in complex low-altitude environments. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor. In the drawings: Figure 1 is a flowchart of a low false alarm recognition method for unmanned aerial vehicles based on radar multi-dimensional micro-motion features provided by the embodiments of the application; Figure 2An unmanned aerial vehicle low false alarm identification whole process schematic diagram provided by an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to enable personnel in the technical field to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor should fall within the scope of protection of the present application.

[0020] It should be noted that the present application includes three parts: low-altitude target radar micro-Doppler feature extraction, low-altitude target identification based on micro-motion characteristics, and identification result timing optimization based on dynamic sliding window. In the feature extraction part, the present application is based on the CFAR constant false alarm detection algorithm, constructs a distance-Doppler domain two-level detection mechanism, and extracts the effective extraction of the low-altitude target multi-dimensional micro-Doppler discrimination feature based on the detection result. In the target identification part, the present application proposes a low-altitude target identification method based on multi-level threshold discrimination according to the extracted multi-dimensional radar features, and realizes the robust identification of unmanned aerial vehicle and non-unmanned aerial vehicle targets. In the identification optimization part, the present application proposes a dynamic sliding window identification result optimization method, which realizes the timing fusion and context perception correction of the identification result of each track point at the track level, thereby effectively improving the continuity and reliability of the identification result.

[0021] Meanwhile, based on the unmanned aerial vehicle target low false alarm identification method based on radar multi-dimensional micro-motion features of the present application, a radar needs to be provided in advance to continuously observe the low-altitude area to obtain target echo electromagnetic signals. Then, the present application extracts target multi-dimensional micro-Doppler discrimination features from the current track point echo based on the distance-Doppler two-level detection method, and constructs multi-level discrimination thresholds based on the differential micro-motion characteristics of low-altitude targets, to realize the robust identification of unmanned aerial vehicle and non-unmanned aerial vehicle targets. Finally, the present application uses a dynamic sliding window to realize the timing fusion and identification result correction of the identification result at the track level, effectively improves the identification consistency and false alarm suppression ability, and finally realizes the high-confidence and low false alarm identification of unmanned aerial vehicle targets in a complex low-altitude environment.

[0022] An unmanned aerial vehicle low false alarm identification method based on radar multi-dimensional micro-motion features is provided in an embodiment of the present application, as shown in Figure 1 The unmanned aerial vehicle low false alarm identification method based on radar multi-dimensional micro-motion features specifically includes steps S101-S104: S101. Based on the target's micro-motion Doppler spectrum, multi-dimensional features are extracted and processed between the target's main signal and the micro-motion signal at low altitude to obtain the target's multi-dimensional micro-motion features. Among them, the target's multi-dimensional micro-motion features include: harmonic interval features, harmonic isolation features, and main harmonic ratio features.

[0023] Specifically, it is necessary to first use a CFAR detector with a low false alarm rate to perform target detection processing on the range-Doppler plane of the current track point echo to obtain the first target detection result.

[0024] Furthermore, based on the Doppler domain and through a secondary CFAR detector, the target's micro-motion harmonic signal in the current track point is processed to obtain the second target detection result.

[0025] Furthermore, based on the first target detection result and the second target detection result, the target is subjected to detection fusion extraction processing between the target main signal and the micro-motion signal to obtain the target Doppler domain detection result.

[0026] In one embodiment, Figure 2 This application provides a schematic diagram of the entire process for identifying low false alarm rates in unmanned aerial vehicles (UAVs) in an embodiment of the present application. Figure 2 As shown, low-altitude targets typically generate lift and propulsion through micro-motion behaviors such as rotor rotation and wing flapping. Different micro-motion modes correspond to significantly different Doppler spectral spread characteristics. To extract discriminative Doppler spectral features, this application employs a two-stage detection process: First, a CFAR detector with a low false alarm rate performs target detection on the range-Doppler (RD) plane of the current trackpoint echo, obtaining a first target detection result; second, a second-stage CFAR detector detects the target's micro-motion harmonic signals in the Doppler domain, obtaining a second target detection result; finally, based on the two-stage detection results, that is, by comprehensively extracting the target's micro-motion features from the target's Doppler domain detection results.

[0027] Furthermore, based on the target Doppler domain detection results and the target's micro-motion Doppler spectrum, the harmonic with the smallest interval to the main Doppler frequency must be selected. The difference between the smallest harmonic and the main Doppler frequency is then calculated to obtain the fundamental frequency. Based on the fundamental frequency, a search process is performed within a preset absolute frequency error tolerance to detect and obtain several higher harmonic points, which are then included in the harmonic set. Finally, the harmonic set and the main Doppler frequency are combined to obtain the target's harmonic interval characteristics.

[0028] In one embodiment, such as Figure 2 As shown, in harmonic interval feature extraction, it is necessary to first select the harmonic with the smallest interval to the main Doppler frequency based on the target Doppler domain detection results, and then calculate the difference between its frequency and the main Doppler frequency. ,by As the fundamental frequency, within the set absolute frequency error tolerance... The system searches for the presence of higher harmonics. Detected higher harmonic points are included in the harmonic set and, together with the main Doppler, constitute the harmonic interval characteristics of the target. If... If a sufficient number of harmonics are not found at the fundamental frequency (3-5 is an empirical value, and the specific quantitative value can be adjusted according to the actual project), then the second smallest interval is used instead. As the fundamental frequency, repeat the above process until all candidate fundamental frequencies have been traversed.

[0029] Furthermore, based on the harmonic interval characteristics, a pair of adjacent harmonic points of the target are first located. Then, the average power of the pair of adjacent harmonic points is calculated and determined as the harmonic power. Next, the average power of the adjacent signals in the pair of adjacent harmonic points is calculated and determined as the noise floor power. The difference between the harmonic power and the noise floor power is also calculated to obtain the energy focusing metric of the pair of adjacent harmonic points. Finally, all adjacent harmonic pairs are traversed to obtain the set of all power difference values ​​based on different energy focusing metrics. The set of all power difference values ​​is then determined as the harmonic isolation characteristic of the target.

[0030] In one embodiment, such as Figure 2 As shown, in the harmonic isolation feature extraction, firstly, using the extracted harmonic interval features as priors, a pair of adjacent harmonic points are located. Secondly, the average power of this pair of harmonics is calculated and used as the harmonic power. Then, the average power of the signal between two adjacent harmonics is calculated and used as the noise floor power. Further, the difference between the harmonic power and the noise floor power is calculated as the energy focusing measure of this pair of harmonic signals. Finally, all adjacent harmonic pairs are traversed to obtain the set of all power difference values, which constitute the harmonic isolation feature.

[0031] Furthermore, the principal harmonic ratio is first obtained based on the amplitude decibel values ​​of each harmonic in the harmonic interval characteristics. That is, the amplitude decibel values ​​of each harmonic in the harmonic interval characteristics need to be obtained first. Then, the amplitudes of each harmonic in the harmonic interval characteristics are numerically compared, and the harmonic with the largest amplitude is determined as the representative harmonic of the micro-Doppler. Finally, the difference between the target principal Doppler and the maximum harmonic amplitude decibel value of the representative harmonic is calculated to obtain the principal harmonic ratio characteristics.

[0032] In one embodiment, such as Figure 2 As shown, in the extraction of principal harmonic ratio features, firstly, the amplitude decibel value of each harmonic is obtained based on the harmonic interval feature; secondly, the amplitude of each harmonic of the target is compared, and the harmonic with the largest amplitude is determined as the representative harmonic of micro-Doppler; finally, the difference between the target principal Doppler and the amplitude decibel value of the largest harmonic in the representative harmonic is calculated, and this is used as the principal harmonic ratio feature.

[0033] Furthermore, the harmonic interval characteristics, harmonic isolation characteristics, and principal harmonic ratio characteristics are combined in a multidimensional manner to obtain the target multidimensional micro-motion characteristics.

[0034] S102. Based on the micro-motion characteristics of low-altitude targets, multi-level discrimination thresholds are set for the multi-dimensional micro-motion features of the targets, and these thresholds are then connected in series to obtain a series three-level threshold judgment mechanism. That is, multi-level discrimination thresholds are constructed based on the micro-motion characteristics of low-altitude targets to achieve the classification and identification of UAVs and non-UAV targets. Discrimination rules are established for the three micro-motion features: harmonic interval characteristics, harmonic isolation characteristics, and principal harmonic ratio characteristics, forming a series three-level threshold judgment mechanism.

[0035] Specifically, based on the UAV micro-motion model, the target's Doppler spectrum is first processed to determine whether it contains a preset number of harmonics to classify the target. Then, based on the first determination rule, a first-level threshold determination mechanism based on harmonic interval characteristics is constructed. The first-level threshold determination mechanism includes: potential UAVs and non-UAVs.

[0036] In one embodiment, such as Figure 2 As shown, a first-level threshold for discrimination is set based on the harmonic interval characteristics. According to the UAV micro-motion model, its Doppler spectrum should exhibit the characteristic of having the main Doppler as the center and the harmonics equally distributed on both sides of the main Doppler. Therefore, this application will use whether the target spectrum contains a sufficient number of harmonics as the initial criterion for judging the target category. The first judgment rule is as follows: ,in, For the target category, Indicates the number of harmonics detected. This is the characteristic threshold for harmonic intervals.

[0037] Furthermore, based on the high signal-to-noise ratio and narrow bandwidth characteristics of the harmonic components of UAVs, threshold judgments are made on the target's harmonic signals under the relevant signal-to-noise ratio and bandwidth narrowness. Based on the second judgment rule, a two-level threshold judgment mechanism based on harmonic isolation characteristics is constructed. The two-level threshold judgment mechanism includes: potential UAVs and non-UAVs.

[0038] In one embodiment, a second-level threshold is set based on the harmonic solitary characteristics. Ideally, the harmonic components of the UAV should exhibit high signal-to-noise ratio and narrowband characteristics, typically showing a narrowband Sinc spread pattern in simulations and actual measurements. This application uses whether the harmonic signal possesses sufficiently high power salience (i.e., sufficiently high signal-to-noise ratio) and sufficiently narrow bandwidth as the second criterion, namely: ,in This represents the difference in decibels between the harmonic point power and the inter-harmonic noise floor power. This is the threshold for harmonic isolation characteristics.

[0039] Furthermore, based on the radar scattering interface of the UAV fuselage, the power ratio between the target's principal Doppler component and the strongest harmonic component must be calculated to obtain the minimum principal harmonic ratio.

[0040] Furthermore, based on the threshold judgment condition of minimum principal harmonic ratio and the third judgment rule, a three-level threshold judgment mechanism based on principal harmonic ratio characteristics is obtained. The three-level threshold judgment mechanism includes: potential drones and non-drones.

[0041] In one embodiment, a threshold is set based on the principal harmonic ratio characteristics. The radar cross-section (RCS) of the UAV fuselage is typically much larger than that of the rotor blades, causing the power of the target's principal Doppler component to be significantly higher than the power of the harmonic components. Therefore, this application uses whether the target's minimum principal harmonic ratio (i.e., the power ratio of the principal Doppler to the strongest harmonic) reaches a certain threshold as a criterion, and the third criterion is: ,in, The target is the main Doppler-strongest harmonic power ratio. The main harmonic ratio characteristic threshold.

[0042] Furthermore, the first-level threshold determination mechanism, the second-level threshold determination mechanism, and the third-level threshold determination mechanism are sequentially connected in series to obtain a series three-level threshold determination mechanism.

[0043] S103. Based on the threshold determination result of the serial three-level threshold determination mechanism, the target identification result under the target track is obtained.

[0044] Specifically, a series of three-level threshold judgment mechanisms are used to sequentially judge the multi-dimensional micro-motion characteristics of the target.

[0045] Furthermore, if each threshold determination mechanism involves potential drones, the threshold determination result is determined as the drone identification result. If each threshold determination mechanism involves non-drones, the threshold determination result is determined as the non-drone identification result.

[0046] Furthermore, based on the UAV identification results and the non-UAV discrimination results, the target identification result under the target track is obtained. That is, the above three thresholds need to be used in series. When all three are satisfied, the target is determined to be a UAV; otherwise, it is a non-UAV, thereby achieving low false alarm recognition of UAV targets.

[0047] S104. Through a preset dynamic sliding window, the recognition results of individual targets in the track level are subjected to temporal fusion and target recognition optimization processing to obtain the optimized target recognition results in the same target track.

[0048] It should be noted that, in order to improve the temporal consistency and reliability of the identification results and suppress false alarms and missed detections caused by noise, clutter, or tracking failures in the waypoint identification results, this application proposes a waypoint hierarchical identification optimization method based on dynamic sliding windows. This mechanism does not rely on the traditional "fixed input-fixed output" sliding window mode, but instead achieves dynamic fusion and real-time correction of historical results within the window on the identification result sequence of the same waypoint.

[0049] Specifically, firstly, a sliding window is used to control the sliding strategy of each target identification result in the same track level. Then, the historical sliding window optimization results in the dynamic sliding window are used as the reference value within the window.

[0050] Furthermore, the in-window reference values ​​are used to recursively update the identification results of individual targets within the same target track. Based on historical sliding window optimization results, consistency constraints are applied to the target identification results of the current track to correct abnormal fluctuations in the identification results of individual targets, thus obtaining optimized target identification results for each track point within the same target track.

[0051] In one embodiment, consider a line containing The target trajectory of each tracking point , of which, at each moment Each corresponds to a target recognition result obtained in step S103. Traditional sliding window methods typically apply a fixed-length window to the original recognition sequence, with the reference value within the window serving as the initial recognition result. However, the dynamic sliding window strategy proposed in this application uses historical sliding window optimization results as reference values ​​within the window during the window sliding process to evaluate the identification results of the current waypoint, thereby achieving a recursive update of the results within the window.

[0052] As a feasible implementation method, in a specific implementation method for optimizing target recognition results, it can be based on The fusion judgment condition is obtained. .in, It is a dynamic sliding window; For reference window The weighting factors, in the absence of prior information, can be set as follows: ; for The historical target identification results in the historical sliding window optimization results at the given time; t is the time corresponding to the track point in the same target track. Then, based on... The fusion decision function is obtained. .in, For historical target identification results, and Finally, let the length of the sliding window be... ,but The original identification results of the time track points are The optimized recognition result after dynamic sliding window is the optimized target recognition result. In other words: according to Optimized target recognition results are obtained. .in, This represents the original target identification results at the track level.

[0053] As a feasible implementation method, the core of the aforementioned sliding window optimization mechanism lies in using the consistency constraint of historical identification results within the window to correct the current track point identification result. When problems such as radar front-end detection misses, tracking beam skew, or measurement miscorrelation cause abnormal fluctuations in the current track point identification result, this mechanism can correct it based on the optimized and stable identification results within the window, thereby effectively suppressing jumps in identification results on the same target track and improving the robustness of the overall system performance.

[0054] In one embodiment, a low-altitude target detection scenario using a single radar can be used as an example to illustrate and analyze the aforementioned low-false alarm rate (False Alarm) identification algorithm for UAVs based on multi-dimensional micro-motion characteristics: After setting up an X-band ground-based radar in a relatively open area, the radar's operating mode was adjusted to Track and Search (TAS) mode, and continuous observation was conducted over the low-altitude area. A total of 10 batches of target tracks were collected, including 2 batches of UAV target data (cooperative targets) and 8 batches of measurement data of natural targets (bird flocks, non-cooperative targets). The UAV targets included two small quadcopter UAVs, DJI Phantom and DJI Inspire, which were piloted by UAV operators and flew radially back and forth within a range of 2-3 km from the radar. The collected bird flock targets were active within a range of 5 km from the radar.

[0055] Table 1. Verification results of the target recognition method based on actual test data.

[0056] The recognition results obtained after experimental verification of the low-altitude target recognition algorithm proposed in this application are shown in Table 1. For the DJI Phantom drone collected in cooperation, the recognition rate was 83.42% before the sliding window optimization and increased to 99.85% after optimization; for the DJI Inspire drone, the recognition rate was 86.47% before optimization and reached 97.40% after optimization. In terms of false alarm control, the false alarm rate was 0.10% before optimization and decreased to 0.00% after sliding window optimization. The experimental results show that the low-altitude target radar recognition algorithm proposed in this application can control the false alarm rate to an extremely low level while maintaining the robust target recognition capability of drones, and has good practical value.

[0057] In addition, this application also provides a low false alarm rate identification device for unmanned aerial vehicles (UAVs) based on multi-dimensional micro-motion features. The low false alarm rate identification device for UAVs based on multi-dimensional micro-motion features specifically includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform: Based on the Doppler spectrum of the target micro-motion, multi-dimensional features are extracted and processed between the target main signal and the micro-motion signal at low altitude to obtain the multi-dimensional micro-motion features of the target; among them, the multi-dimensional micro-motion features of the target include: harmonic interval features, harmonic isolation features, and main harmonic ratio features; Based on the micro-motion characteristics of low-altitude targets, the multi-dimensional micro-motion features of the targets are processed by setting multi-level discrimination thresholds, and then connected in series to obtain a series three-level threshold judgment mechanism. Based on the threshold determination result of the serial three-level threshold determination mechanism, the target identification result under the target track is obtained; By using a preset dynamic sliding window, the recognition results of individual targets in the track level are fused in time and optimized to obtain the optimized target recognition results in the same target track.

[0058] This application's embodiments achieve high accuracy and low false alarm rate identification of low-altitude UAV targets through low-false alarm rate identification based on radar multi-dimensional micro-motion characteristics. It effectively adapts to changes in target motion states and target types, significantly improving the ability to distinguish between UAVs and non-UAV targets. Especially when facing targets outside the database, and because the proposed identification based on target micro-physical characteristics does not rely on a data-driven model, it effectively reduces the dependence of the identification method on sample data. Furthermore, through a dynamic sliding window optimization strategy, the temporal consistency of identification results can be enhanced at the track level, improving the system's reliability and practicality in complex low-altitude environments.

[0059] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0060] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0061] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0062] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0063] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0064] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0065] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0066] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0067] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0068] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0069] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of this specification.

Claims

1. A method for low false alarm recognition of a UAV based on radar multi-dimensional micro-motion features, characterized in that, The method comprises: According to the target micro-doppler spectrum, the multi-dimensional feature extraction processing between the target main body signal and the micro-motion signal under low altitude is carried out to obtain the target multi-dimensional micro-motion feature; wherein the target multi-dimensional micro-motion feature includes: harmonic interval feature, harmonic isolation feature and main harmonic ratio feature; According to the low-altitude target micro-motion characteristics, the target multi-dimensional micro-motion feature is subjected to multi-stage discriminant threshold setting processing to obtain a series of three-stage threshold discriminant mechanism; Based on the threshold discriminant result of the series of three-stage threshold discriminant mechanism, the target recognition result under the target track is obtained; Through the preset dynamic sliding window, the single target recognition result in the track level is subjected to time sequence fusion and target recognition optimization processing to obtain the optimized target recognition result in the same target track.

2. The method of claim 1, wherein the method comprises: Before the multi-dimensional feature extraction processing between the target main body signal and the micro-motion signal under low altitude according to the target micro-doppler spectrum is carried out to obtain the target multi-dimensional micro-motion feature, the method further comprises: Through the low false alarm rate CFAR detector, target detection processing is carried out on the range-doppler plane of the current track point echo to obtain the first target detection result; Based on the Doppler domain, and through the two-stage CFAR detector, the target in the current track point is subjected to target micro-motion harmonic signal detection processing to obtain the second target detection result; According to the first target detection result and the second target detection result, the target is subjected to detection fusion extraction processing about the target main body signal and the micro-motion signal to obtain the target Doppler domain detection result.

3. The method of claim 2, wherein the method further comprises: According to the target micro-doppler spectrum, the multi-dimensional feature extraction processing between the target main body signal and the micro-motion signal under low altitude is carried out to obtain the target multi-dimensional micro-motion feature, specifically comprising: According to the target Doppler domain detection result and based on the target micro-doppler spectrum, the harmonic with the smallest interval from the main Doppler frequency is selected; and the difference between the smallest harmonic and the main Doppler frequency is calculated to obtain the fundamental frequency; Based on the fundamental frequency, search processing is carried out on whether there is a high-order harmonic in the preset absolute frequency error tolerance to detect and obtain a plurality of high-order harmonic points, and the high-order harmonic points are included in the harmonic set; The harmonic set and the main Doppler are subjected to common composition processing to obtain the harmonic interval feature of the target; Based on the harmonic interval feature, a pair of adjacent harmonic points of the target are located; The average power of the pair of adjacent harmonic points is calculated and determined as the harmonic power; The average power of the adjacent signals in the pair of adjacent harmonic points is calculated and determined as the noise floor power; The difference between the harmonic power and the noise floor power is calculated to obtain the energy focusing metric of the pair of adjacent harmonic points; All adjacent harmonic pairs are traversed to obtain a set of all power difference values based on different energy focusing metrics; and the set of all power difference values is determined as the harmonic isolation feature of the target; Based on the amplitude decibel value of each harmonic in the harmonic interval feature, the main harmonic ratio feature is obtained; The harmonic interval feature, the harmonic isolation feature and the main harmonic ratio feature are subjected to multi-dimensional feature combination to obtain the target multi-dimensional micro-motion feature.

4. The method of claim 3, wherein the method further comprises: The main harmonic ratio feature is obtained based on the amplitude decibel values of each harmonic in the harmonic interval feature, and specifically includes: Obtaining the amplitude decibel values of each harmonic in the harmonic interval feature; Numerical comparison is performed on the amplitudes of each harmonic in the harmonic interval feature, and the harmonic with the largest amplitude is determined as the representative harmonic of the micro-Doppler; Difference calculation is performed between the target main Doppler and the maximum harmonic amplitude decibel value of the representative harmonic to obtain the main harmonic ratio feature.

5. The method of claim 1, wherein the method further comprises: According to the low-altitude target micro-motion characteristics, the target multi-dimensional micro-motion characteristics are subjected to multi-stage threshold setting processing, and a series of three-stage threshold determination mechanisms are obtained in series, specifically including: According to the UAV micro-motion model, whether the target Doppler spectrum contains a preset number of harmonics is determined as the target category, and based on the first determination rule, a first-stage threshold determination mechanism based on the harmonic interval feature is constructed; wherein the first-stage threshold determination mechanism includes: potential UAV and non-UAV; According to the high signal-to-noise ratio and narrow band characteristics of the harmonic components of the UAV, threshold judgment is performed on the harmonic signal of the target under the conditions of high signal-to-noise ratio and narrow bandwidth, and based on the second determination rule, a second-stage threshold determination mechanism based on the harmonic isolation feature is constructed; wherein the second-stage threshold determination mechanism includes: potential UAV and non-UAV; According to the radar scattering interface of the UAV body, power ratio calculation is performed between the target main Doppler component and the strongest harmonic component of the target to obtain the minimum main harmonic ratio; Based on the threshold judgment condition of the minimum main harmonic ratio and the third determination rule, a three-stage threshold determination mechanism based on the main harmonic ratio feature is obtained; wherein the three-stage threshold determination mechanism includes: potential UAV and non-UAV; The first-stage threshold determination mechanism, the second-stage threshold determination mechanism, and the three-stage threshold determination mechanism are sequentially subjected to threshold series processing to obtain the series of three-stage threshold determination mechanisms.

6. The method of claim 1, wherein the method further comprises: Based on the threshold determination result of the series of three-stage threshold determination mechanisms, the target recognition result under the target track is obtained, specifically including: Through the series of three-stage threshold determination mechanisms, the target multi-dimensional micro-motion characteristics of the target are sequentially subjected to threshold judgment; If the potential UAV is in each threshold determination mechanism, the threshold determination result is determined as the UAV recognition result; If there is a non-UAV in each threshold determination mechanism, the threshold determination result is determined as the non-UAV discrimination result; Based on the UAV recognition result and the non-UAV discrimination result, the target recognition result under the target track is obtained.

7. The method of claim 1, wherein the method further comprises: Through the preset dynamic sliding window, the single target recognition result in the track level is subjected to time sequence fusion and target recognition optimization processing to obtain the optimized target recognition result in the same target track, specifically including: Through the dynamic sliding window, each target recognition result in the same track level is subjected to sliding strategy control; The historical sliding window optimization result in the dynamic sliding window is taken as the window reference value; The single target recognition result in the same target track is recursively identified and updated through the reference value in the window, and the consistency constraint is performed on the current track target recognition result based on the historical sliding window optimization result, so that the abnormal fluctuation correction processing of the single target recognition result is completed, and the optimization target recognition result of each track point in the same target track is obtained.

8. The method according to claim 7, wherein, according to The fusion judgment condition is obtained. ;in, For the aforementioned dynamic sliding window; For reference window Weighting factors; for The historical target identification result in the historical sliding window optimization results at the given time; t is the time corresponding to the track point in the same target track; According to , a fusion decision function is obtained; wherein the is the historical target recognition result, and ; According to , the optimization target recognition result is obtained; wherein, is the original target recognition result in the track level.

Images