Morphology-based scene surveillance radar signal processing method

Abstract

The application discloses a scene monitoring radar signal processing method based on morphology, which comprises the following steps: S1, acquiring a scene monitoring radar video signal; S2, obtaining a B mode display graph by displaying the radar video signal in a B mode; S3, obtaining a clutter graph by a preset algorithm; S4, performing binary graph processing on the B mode display graph and the clutter graph to obtain a B mode foreground graph; S5, obtaining a B mode binary image contour set according to the B mode foreground graph, and converting the B mode binary image contour set into a P mode suspected target set through transformation; S6, obtaining a target set under P mode display according to the P mode suspected target set; and S7, performing target tracking on the target set under P mode display by using a preset data association algorithm to obtain track information of a real target. The application utilizes an integrated mode of target detection, target identification and multi-target tracking, effectively suppresses the clutter and false targets, and reduces the false alarm rate.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing, and in particular to a morphology-based method for processing scene surveillance radar signals. Background Technology

[0002] With the development of my country's airport scale, more and more airports are deploying and using surface surveillance radar to assist air traffic control centers in airport dispatching. As of the end of 2021, more than 20 airports in my country had installed and were using airport surface surveillance radar, mainly distributed along the southeast coast and in provinces with high air traffic volume such as Heilongjiang, Jilin, Yunnan, and Xinjiang. In the long term, in the Civil Aviation Administration of my country's plan for domestic civil aviation airport surveillance systems, airport surface surveillance radar serves as an auxiliary and supplementary means to the Automatic Dependent Surveillance-Broadcast (ADS-B) system. Compared with other monitoring methods such as ADS-B and transponder-based multi-base station positioning systems, airport surface surveillance radar has advantages such as not requiring target cooperation, long operating range, high data rate, and all-weather and 24-hour operation. Since the 1960s, foreign countries have begun using surface surveillance radar for airport ground monitoring. Currently, major international manufacturers of surface movement radar (SMR) include Thales (France), Terma (Denmark), Indra (Spain), and Selex (USA). The surface surveillance radar signal processing systems from the above companies are priced relatively high. Domestic customers have strong customization demands for signal processing equipment, while foreign equipment systems suffer from excessively high costs associated with responding to customer changes. Furthermore, the use of numerous traditional radar signal processing parameters necessitates high levels of technical support, hindering system maintenance. Inconsistencies between domestic and foreign equipment result in weak system portability and high communication costs; moreover, for manufacturers of A-SMGCS and integrated control tower systems, this leads to difficulties in integration and after-sales maintenance. Therefore, it is imperative to conduct independent research and development of airport surface surveillance radar signal processing systems, and to research and manufacture surface surveillance radars with independent intellectual property rights, superior performance, and reliable operation.

[0003] Traditional airport surface surveillance radar signal processing has the following significant weaknesses:

[0004] 1. Traditional target detection using one-dimensional signal extraction typically requires high computation speed and is usually implemented in real-time hardware, typically on an FPGA. Therefore, it is less flexible and more expensive than software radio architecture.

[0005] 2. Traditional field radar signal processing uses one-dimensional constant false alarm detection technology for target detection. This method has acceptable target detection capability in the range direction, but when combined with angle signals to integrate into two-dimensional target detection, under two-dimensional coupling, due to the false alarm threshold setting, it is easy to miss targets. In the case of multiple targets obscuring, the clutter edge performance is also poor, and in the process of detecting slow targets, self-obscuring phenomenon is easy to occur, leading to missed targets.

[0006] 3. When the target energy and clutter / noise levels are comparable (low signal-to-clutter / noise ratio), the detector may experience missed detections or false alarms, requiring processing over a longer timescale through tracking. This process mainly relies on the motion correlation between target frames and the lack of correlation between clutter / noise frames, for example, by forming track hypotheses to solve the data correlation problem;

[0007] 4. Because existing surface surveillance radars have high range and angle resolution, they are already able to initially identify the type of target through image recognition technology.

[0008] The reasons for the above-mentioned shortcomings are:

[0009] 1. Conventional field surveillance radar signal processing mostly uses one-dimensional signal processing technology, which is difficult to upgrade from the perspective of software radio. Modifying the algorithm requires more hardware support, which in turn leads to more cost investment.

[0010] 2. There is little involvement in two-dimensional signal monitoring and extraction, and few techniques for signal extraction are studied from the perspective of numerical image processing; in addition to the amplitude information of the signal, there are also contour information, layer information, bit depth information, size information, etc. of two-dimensional signals that are difficult to extract using traditional methods, and this information is discarded.

[0011] 3. Traditional methods do not fully utilize the information obtained from the two-dimensional high-resolution real aperture images detected by the surface surveillance radar, resulting in data overload in target tracking and causing target tracking to be lost. Summary of the Invention

[0012] To address the shortcomings of existing technologies, this invention provides a morphology-based field surveillance radar signal processing method that effectively suppresses clutter and false targets and reduces the false alarm rate by integrating target detection, target recognition, and multi-target tracking.

[0013] This invention discloses a morphology-based method for processing scene surveillance radar signals, comprising the following steps:

[0014] S1, acquire scene surveillance radar video signals;

[0015] S2, The radar video signal is displayed in B mode to obtain a B mode display image;

[0016] S3, based on the B mode display image, obtain the dynamic clutter map and the offline setting area clutter map by averaging through a sliding window, and obtain the clutter map based on the dynamic clutter map and the area clutter map by a preset algorithm;

[0017] S4. Perform binary image processing on the B-mode display image and clutter image to obtain the B-mode foreground image;

[0018] S5, obtain the B-mode binary image contour set based on the B-mode foreground image, and transform the B-mode binary image contour set into the P-mode suspected target set through transformation;

[0019] S6, process the suspected target set in P mode to obtain the target set displayed in P mode;

[0020] S7 uses a preset data association algorithm to track the target set displayed in P mode and obtain the trajectory information of the real target.

[0021] Preferably, step S3 specifically includes:

[0022] The B-mode display image is stored in the accumulation pool. It is determined whether the number of accumulations in the accumulation pool is greater than N. If not, Gaussian filtering is applied to the B-mode display image in the accumulation pool, and dynamic sliding window accumulation is performed to obtain a dynamic clutter image.

[0023] Based on the dynamic clutter map, K-frame images are accumulated and offline settings are performed to obtain a regional clutter map;

[0024] Based on the dynamic clutter map and the corresponding maximum value of the regional clutter map, a clutter map is obtained.

[0025] Preferably, if the number of accumulated elements in the accumulation pool is greater than N, then the earliest element in the queue is deleted and the accumulation pool is updated.

[0026] Preferably, step S6 specifically includes:

[0027] For the set of suspected targets in P mode, if the target size meets the interval threshold, the target area meets the interval threshold, and the target can be identified, then the target is determined to be a suspected target; all suspected targets are stored to obtain the target set displayed in P mode.

[0028] Preferably, if the target size does not meet the interval threshold, the target is determined to be a false target.

[0029] Preferably, if the area of ​​the target region does not meet the interval threshold, the target is determined to be a false target.

[0030] Preferably, if a target cannot be identified, it is determined to be a false target.

[0031] Preferably, the preset data association algorithm includes the GNN data association algorithm.

[0032] The beneficial effects of this invention are as follows:

[0033] This invention starts with the basic framework of tracking and detection, and designs a field surveillance radar signal processing technology to achieve uninterrupted processing of field surveillance signals. It also has low hardware resource requirements, enabling the implementation of software-defined radio functionality on low-performance hardware platforms. The main features include:

[0034] 1. This invention enables the use of multiple frames of observation data for long-term noncoherent accumulation, thereby solving the problem of weak target detection;

[0035] 2. This invention can realize the integrated functions of target detection, clutter map accumulation, tracking and identification of surface surveillance radar signals. It can identify low-resolution aircraft targets on the runway and can identify targets and preliminarily identify aircraft through target classification technology.

[0036] 3. This invention can retain all the original data, and has low computational, data storage and transmission overhead; and the accumulation process has low requirements for the accuracy of the target prior knowledge. Attached Figure Description

[0037] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0038] Figure 1 A schematic flowchart of a morphology-based field surveillance radar signal processing method provided in an embodiment of the present invention;

[0039] Figure 2 This is a flowchart illustrating step S3 provided in an embodiment of the present invention;

[0040] Figure 3 This is a flowchart illustrating step S6 provided in an embodiment of the present invention;

[0041] Figure 4 This is the original video image in B-display mode provided in the embodiments of the present invention;

[0042] Figure 5 This is a set of suspected targets with accompanying shadow areas in the B-mode display provided in this embodiment of the invention;

[0043] Figure 6 This is a set of suspected targets with accompanying shadow areas in the P display mode provided in this embodiment of the invention;

[0044] Figure 7 A schematic diagram of six identified aircraft targets provided in an embodiment of the present invention;

[0045] Figure 8 This invention provides a map of the same target located in four consecutive frames in the communication channel, as provided in an embodiment of the invention.

[0046] Figure 9 The image shows four consecutive frames of the same target located on runway 02R, as provided in this embodiment of the invention. Detailed Implementation

[0047] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0048] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0049] In traditional surface surveillance signal processing, target detection and extraction techniques are based on the conventional Constant False Alarm Rate (CFAR) detection method. This approach typically employs on-chip signal acquisition and extraction, preprocessing via an FPGA on-chip system, and subsequent signal and data processing via a dedicated DSP. However, the echo signals from airport surface surveillance radar differ significantly from those of conventional radar. Traditional radar clutter background and target detection methods, as well as CFAR, suffer from slow target radial velocity and complex target clustering, hindering accurate target detection. While traditional methods have advantages, such as mature signal processing technology and high hardware efficiency, they also suffer from algorithmic complexity, excessive parameters, and limited adaptability. Traditional range-oriented CFAR detection methods, although capable of roughly detecting target blocks, suffer from numerous false alarms and missed alarms within target blocks. Relying on rule-based algorithms to eliminate isolated false alarms, fill voids within targets, and merge target clusters in the azimuth direction presents significant challenges.

[0050] With the rapid development of high-resolution radar technology, airport surface surveillance radar has also made rapid progress. The Terma surface surveillance radar at Chengdu Shuangliu Airport, for example, has achieved a range resolution of 10 ns. High range resolution is achieved by using wide-bandwidth signals, high azimuth resolution is achieved by using narrow beams in the azimuth direction, and high-speed network transmission is used to obtain two-dimensional high-resolution real-aperture images. Therefore, airport surface surveillance radar can now obtain high-resolution images that allow for preliminary identification of target types. In light of this, based on the actual operational needs of surface surveillance radar at large airports in my country, a morphology-based surface surveillance radar signal processing system has been invented. This system provides a new technical means for monitoring control targets and ensuring operational safety, realizing integrated tracking functions from target extraction, target tracking, and target identification for surveillance radar.

[0051] like Figure 1 As shown, this embodiment of the invention provides a morphology-based field surveillance radar signal processing method, including the following steps:

[0052] S1, acquire scene surveillance radar video signals;

[0053] S2, display the radar video signal in B mode to obtain the B mode display diagram;

[0054] S3, based on the B mode display image, obtain the dynamic clutter map and the offline set area clutter map through sliding window averaging, and obtain the clutter map based on the dynamic clutter map and the area clutter map through a preset algorithm;

[0055] S4. Perform binary image processing on the B-mode display image and clutter image to obtain the B-mode foreground image;

[0056] S5. Obtain the B-mode binary image contour set based on the B-mode foreground image, and transform the B-mode binary image contour set into the P-mode suspected target set through transformation.

[0057] S6, process the suspected target set in P mode to obtain the target set displayed in P mode;

[0058] S7 uses a preset data association algorithm to track the target set displayed in P mode and obtain track information.

[0059] like Figure 2 As shown, step S3 specifically includes:

[0060] The B-mode display image is stored in the accumulator pool. It is determined whether the number of accumulations in the accumulator pool is greater than N. If not, the B-mode display image in the accumulator pool is subjected to Gaussian filtering and dynamic sliding window accumulation to obtain the dynamic clutter image.

[0061] A regional clutter map is obtained by accumulating K-frame images based on dynamic clutter maps and performing offline settings.

[0062] The clutter map is obtained by taking the maximum value of the corresponding cell from the dynamic clutter map and the regional clutter map.

[0063] The scene surveillance radar video signal is acquired from the radar. Assume the radar is divided into M range cells (m = 1, 2, ..., M) in the range direction and N range cells (n = 1, 2, ..., N) in the azimuth direction. The clutter map has N × M range-azimuth cells, where (m, n) represents the cell number. The dynamic clutter map is as follows:

[0064]

[0065] Where l is the size of the sliding window. D represents the clutter power of the (n,m)th cell in the single-frame dynamic clutter map formed during the i-th frame scanning period. d (m, n) represents the dynamic clutter power of the (m, n)th unit of the moving average. At this point, Gauss filtering is also used to filter out abnormal spot noise and smooth the image.

[0066] Regional clutter maps are set offline as follows: D s (m, n), this is the offline setting. The principle of the offline setting is to record more than 100 frames of the above dynamic clutter accumulation map under different meteorological conditions in the field. d Accumulate (m, n). Assume the ADC conversion bit depth is x bits, and the sampling point is D at the building location. s (m, n) = 2 x -1, the clutter assignment D for the grassland. S (m, n) = min (Grass (m, n), 2 x -1), where Grass(m,n) can be obtained from the grassland clutter map D within the airport area. d The maximum value is obtained by finding Grass(m, n), i.e., Grass(m, n) = max(D). d (m′, n′)), where (m′, n′) ∈ GrassArea. Typically, the settings are based on a topographic map, and the XY coordinate map is transformed to the RA coordinate map through inverse transformation. Two-dimensional bilinear interpolation can be used for near-end image completion, while at the far end, multiple discrete data are combined by taking the maximum value to merge data units.

[0067] The resulting clutter pattern is: D(m,n) = max(D d (m, n), D s (m, n)). Where: D s (m, n) is static, D d (m, n) is dynamically time-varying.

[0068] In step S3 of this embodiment, the clutter map is generated using non-coherent cumulative sliding window generation. This technology uses 10 sliding windows to generate the clutter map, which can suppress short-period clutter targets. To improve the detection probability, for example, under good weather conditions, the number of sliding windows can be increased. To have a stronger suppression level for dynamic targets, for example, under relatively bad weather conditions, the number of sliding windows can be reduced.

[0069] In this embodiment of the invention, different types of areas are set in conjunction with the airport's plan. For areas such as grass where targets are very unlikely to exist, the target detection threshold in these areas needs to be increased to suppress false alarms. For areas such as buildings where targets are impossible to exist, the target detection threshold in these areas is set to the maximum value (0xFF = 255 for an 8-bit ADC) to completely suppress target generation. For areas such as airport runways and taxiways where targets are highly likely to exist, non-coherent cumulative maps are used to accumulate clutter maps.

[0070] In step S4 of this embodiment of the invention, the extraction of the foreground image in mode B is performed using binary image processing on the mode B display image. Compared to traditional processing on the PPI mode display image, this embodiment of the invention can save graphics storage and improve processing speed.

[0071] like Figure 3 As shown, step S6 specifically includes: for the set of suspected targets in P mode, if the target size meets the interval threshold, the area of ​​the target region meets the interval threshold, and the target can be identified, then the target is determined to be a suspected target; store all suspected targets to obtain the target set under P mode display.

[0072] If the target size does not meet the interval threshold, the target is determined to be a false target.

[0073] If the area of ​​the target region does not meet the interval threshold, the target is determined to be a false target.

[0074] If a target cannot be identified, it is determined to be a false target.

[0075] In this embodiment of the invention, a set of suspected targets in mode P is screened from a set of binary image contours in mode B, combined with an image recognition module, to remove interference signals such as clutter and shadows. Shadow removal is calculated using a combination of morphological and scale methods, and target detection is performed through image classification. Image classification detection involves edge detection and contour extraction of the image, followed by similarity comparison between different contours. A threshold of 0.8 is used; a similarity greater than 0.8 is considered a target, otherwise it is considered a false target. The target needs to be rotated in the image, with the rotation angle equal to the target track angle obtained from multi-target tracking.

[0076] In this embodiment of the invention, the target tracking module adopts the GNN data association algorithm. Compared with conventional association algorithms, this embodiment of the invention, in addition to using kinematic information state variables, also incorporates morphological information, thus enhancing the accuracy of data association and reducing the ID switching phenomenon that reduces target tracking.

[0077] The tracking algorithm here combines polar coordinate measurement with tracking in Cartesian coordinates. Specifically, it uses EKF filtering, GNN data association, and elliptic gates for distance mapping.

[0078]

[0079]

[0080] In a two-dimensional plane, z is the measured value. S is the predicted value of the k-th measurement. k Let be the k-th innovation matrix, χ is used here. -1 A threshold of (0.005.2) = 10.6 is sufficient. Furthermore, a similarity penalty function term is added to the threshold association. The association value is: in P represents the Mahalanobis distance measured between the k-th trajectory i and j-th trajectory; s (i, j) represents the similarity score of the matches, where 0 represents complete dissimilarity and 1 represents complete similarity, i.e., a penalty is applied to the distance calculation. Then, the Munkres or Kuhn assignment algorithm is used for data association.

[0081] The embodiments of the present invention were verified and tested using typical data from Chengdu Shuangliu Airport, wherein... Figure 4 To display the radar video signal in Mode B, a Mode B display image is obtained; a superimposed image of suspected targets is obtained through contour extraction, such as... Figure 5 As shown; after conversion, a set of suspected targets in P display mode can be obtained, such as... Figure 6 As shown; the final track output map can be obtained through identification and tracking, such as... Figure 7 As shown, the track number is above the target, and the target's aircraft identification rate is below; as... Figure 8 , Figure 9 As shown, this is an image of the same target extracted from four consecutive frames at different locations.

[0082] This invention provides a morphology-based method for processing radar signals for scene surveillance. It integrates target detection, target recognition, and multi-target tracking to effectively suppress clutter and false targets, reducing the false alarm rate. The adaptive target extraction method further reduces the missed alarm rate. The integration of target recognition into the scene radar signal processing enhances target detection capabilities. It also has lower requirements for meteorological conditions, reducing the computational load of meteorological clutter maps and lowering computational complexity. Furthermore, this invention improves upon conventional clutter map generation techniques, enhancing target detection capabilities, strengthening multi-target tracking, and reducing the probability of ID swapping.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A morphology-based method for processing scene surveillance radar signals, characterized in that, Including the following steps: S1, acquire scene surveillance radar video signals; S2, The radar video signal is displayed in B mode to obtain a B mode display image; S3, based on the B mode display image, obtain the dynamic clutter map and the offline setting area clutter map by averaging through a sliding window, and obtain the clutter map based on the dynamic clutter map and the area clutter map by a preset algorithm; S4. Perform binary image processing on the B-mode display image and clutter image to obtain the B-mode foreground image; S5, obtain the B-mode binary image contour set based on the B-mode foreground image, and transform the B-mode binary image contour set into the P-mode suspected target set through transformation; S6, process the suspected target set in P mode to obtain the target set displayed in P mode; S7 uses a preset data association algorithm to track the target set displayed in P mode and obtain the trajectory information of the real target.

2. The morphology-based scene surveillance radar signal processing method according to claim 1, characterized in that, Step S3 specifically includes: The B-mode display image is stored in the accumulation pool. It is determined whether the number of accumulations in the accumulation pool is greater than N. If not, Gaussian filtering is applied to the B-mode display image in the accumulation pool, and dynamic sliding window accumulation is performed to obtain a dynamic clutter image. Based on the dynamic clutter map, K-frame images are accumulated and offline settings are performed to obtain a regional clutter map; Based on the dynamic clutter map and the corresponding maximum value of the regional clutter map, a clutter map is obtained.

3. The morphology-based scene surveillance radar signal processing method according to claim 2, characterized in that, If the number of accumulated elements in the accumulation pool is greater than N, then delete the earliest element in the queue and update the accumulation pool.

4. The morphology-based scene surveillance radar signal processing method according to claim 2, characterized in that, Step S6 specifically includes: For the set of suspected targets in P mode, if the target size meets the interval threshold, the target area meets the interval threshold, and the target can be identified, then the target is determined to be a suspected target; all suspected targets are stored to obtain the set of suspected targets displayed in P mode.

5. The morphology-based scene surveillance radar signal processing method according to claim 4, characterized in that, If the target size does not meet the interval threshold, the target is determined to be a false target.

6. The morphology-based scene surveillance radar signal processing method according to claim 4, characterized in that, If the area of ​​the target region does not meet the interval threshold, the target is determined to be a false target.

7. The morphology-based scene surveillance radar signal processing method according to claim 4, characterized in that, If a target cannot be identified, it is determined to be a false target.

8. The morphology-based scene surveillance radar signal processing method according to claim 1, characterized in that, The preset data association algorithm includes the GNN data association algorithm.

Images