Urban low-altitude multi-radar networking method and system based on neural network compensation

Abstract

The invention provides an urban low-altitude multi-radar networking method and system based on neural network compensation, and the method comprises the steps: considering the shielding of urban trees and buildings, and dividing the deployment of a radar detection subsystem into two aspects: an airspace and the ground; and then the flight paths of the radars are fused to form global unmanned aerial vehicle state position and situation information based on a neural network compensation method, and different radars are deployed according to actual terrains and needs in a high-low hierarchical combination mode, so that the cost is saved, non-blind-area detection of urban low-altitude unmanned aerial vehicles is realized, and the detection accuracy is improved. And accurate and stable unmanned aerial vehicle position and situation information is provided for an urban unmanned aerial vehicle detection system.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-system radar networking, and in particular to a multi-radar networking method and system based on neural network compensation at low altitude in a city. Background Art

[0002] In order to promote the economic development of low-altitude airspace, the country has gradually opened up low-altitude airspace since 2020. During this period, the number of drones has increased exponentially and played an increasingly critical role in emerging fields such as air commuting, logistics, tourism and management. However, although drones bring convenience to life, there is also a risk of abuse. In this case, efficient and accurate urban low-altitude drone detection will become one of the core of smart city development.

[0003] Drone radar reflection cross-sectional area is small and maneuverable. Especially when flying in low-altitude urban environment, due to complex terrain and more electromagnetic interference, its detection and identification has always been an internationally recognized technical problem. At present, the commonly used urban drone detection system mainly consists of three detection subsystems: radar, spectrum and optoelectronics. The system first detects and tracks drones in the monitoring area through radar equipment, then guides distributed spectrum passive monitoring equipment to monitor the communication signals of drones, and then guides optoelectronics to identify drones at medium and short distances. Therefore, how to deploy the radar detection subsystem in a complex urban background to achieve full coverage of the monitoring area and provide accurate location information of drones to the other two subsystems is a major problem that the system needs to overcome. Although the system has a good detection effect on urban drones, the system has many devices, is difficult to install and deploy, and is very expensive. Therefore, how to deploy the radar detection subsystem under the premise of saving costs is also a major factor that the system needs to consider.

[0004] At present, different radar systems can play a good monitoring role on drones in corresponding specific scenarios. For example, the three-coordinate low-altitude surveillance radar with conventional azimuth scanning and pitch scanning can accurately detect drones at low altitudes above tens of meters with less interference and shielding; the frequency modulated continuous wave (FMCW) radar system has unique advantages such as low cost, low power consumption, high distance resolution, strong anti-interference ability, no distance blind spot and high data rate, and can continuously detect drones in low-altitude and ultra-low-altitude environments. However, the target echo information of the former is unstable at low altitudes below tens of meters due to the shielding of trees and buildings; the latter has a high radar frequency and large propagation loss of electromagnetic waves, which limits the detection distance of the radar, usually only hundreds of meters.

[0005] Therefore, in the complex environment of the city, a single radar system is difficult to meet the needs of full airspace coverage. It is necessary to network radars of different systems for collaborative detection, complement each other's advantages, and connect with each other in the airspace, thus forming a wider airspace coverage. Summary of the invention

[0006] The technical problem to be solved by the present invention is how to realize low-altitude multi-radar networking under the premise of saving costs in a complex urban environment.

[0007] The present invention solves the above technical problems by the following technical means:

[0008] The multi-radar networking method based on neural network compensation in urban low altitude includes the following steps:

[0009] The urban low altitude is divided into two detection spaces, airspace and ground. Multiple surveillance radars of different systems are deployed in the two detection spaces, and the radar data of different systems are unified in the coordinate system. Then, based on the neural network compensation method, the radar tracks of the same target from different radars are fused to form the global UAV status position and situation information.

[0010] Furthermore, the radar deployment method in the airspace is:

[0011] According to the terrain of the area to be monitored, the approximate scope of the area is obtained;

[0012] An optimization mathematical model is established, which takes minimizing cost and maximizing radar coverage area as the objective function, the number of radars and radar deployment locations as variables, and the total radar budget and surveillance space as constraints; then the number and locations of radars to be deployed are calculated using a genetic optimization algorithm;

[0013] According to the calculated radar position, find a relatively tall building or tower nearby and deploy the radar on the roof of the building or tower.

[0014] Furthermore, the ground radar deployment method is:

[0015] Introduce the geographical environment feature database of the monitored area, study the terrain of the monitored area, and calculate the radar deployment area, which is mainly divided into urban road sections, between buildings and urban open areas;

[0016] For urban roads, low-cost frequency-modulated continuous wave radars are deployed based on the city's basic transportation construction facilities; for buildings, radars with different detection powers are deployed according to the distance between the buildings; for open areas in the city, radars with detection power similar to the area of ​​the area are deployed based on the area.

[0017] Furthermore, the method of unifying the coordinate system of radar data of different systems is as follows:

[0018] A distributed data processing architecture is adopted, and signal processing and data processing modules are embedded in each radar to generate track data independently. The precise longitude and latitude information of each radar site is measured, and then the proj library used for geographic coordinate system conversion and map projection is called to convert the radar track data to the WGS-84 coordinate system. The converted radar track data is transmitted to the data center using the urban traffic public network.

[0019] Furthermore, the method for fusing different radar track data of the same target is as follows:

[0020] Correlate the same target data from different radars: Use a simple joint probability data association method to calculate the Mahalanobis distance d between the track data of the tracked target from different sensors ij And form a distance matrix D, and then calculate the distance threshold value r of the tracking target track data of different radars ij , and each element d in the distance matrix ij Compare with each tracking threshold, and form the judgment results into a confirmation matrix;

[0021] Data fusion is performed on the successfully associated tracks: the fusion center first directly fuses the data output by each different radar, and introduces the radial basis function in the fusion process to effectively reduce the errors that may be caused by interactive covariance. In the training process, some residual errors (ε m (k), m=1,..,n) and the error between the state estimate and the predicted value (X m (k)-ΦX m (k-1),m=1,...,n).

[0022] The present invention also provides a multi-radar networking system based on neural network compensation at low altitude in cities, comprising:

[0023] Airspace and ground radar monitoring module: divides the urban low altitude into two detection spaces: airspace and ground, and deploys multiple surveillance radars of different systems in the two detection spaces;

[0024] Data management module: unify the coordinate system of radar data of different systems;

[0025] Data fusion module: Based on the neural network compensation method, the radar tracks of the same target from different radars are fused to form the global UAV status position and situation information.

[0026] Furthermore, the radar deployment method in the airspace is:

[0027] According to the terrain of the area to be monitored, the approximate scope of the area is obtained;

[0028] An optimization mathematical model is established, which takes minimizing cost and maximizing radar coverage area as the objective function, the number of radars and radar deployment locations as variables, and the total radar budget and surveillance space as constraints; then the number and locations of radars to be deployed are calculated using a genetic optimization algorithm;

[0029] According to the calculated radar position, find a relatively tall building or tower nearby and deploy the radar on the roof of the building or tower.

[0030] Furthermore, the ground radar deployment method is:

[0031] Introduce the geographical environment feature database of the monitored area, study the terrain of the monitored area, and calculate the radar deployment area, which is mainly divided into urban road sections, between buildings and urban open areas;

[0032] For urban roads, low-cost frequency-modulated continuous wave radars are deployed based on the city's basic transportation construction facilities; for buildings, radars with different detection powers are deployed according to the distance between the buildings; for open areas in the city, radars with detection power similar to the area of ​​the area are deployed based on the area.

[0033] Furthermore, the method of unifying the coordinate system of radar data of different systems is as follows:

[0034] A distributed data processing architecture is adopted, and signal processing and data processing modules are embedded in each radar to generate track data independently. The precise longitude and latitude information of each radar site is measured, and then the proj library used for geographic coordinate system conversion and map projection is called to convert the radar track data to the WGS-84 coordinate system. The converted radar track data is transmitted to the data center using the urban traffic public network.

[0035] Furthermore, the method for fusing different radar track data of the same target is as follows:

[0036] Correlate the same target data from different radars: Use a simple joint probability data association method to calculate the Mahalanobis distance d between the track data of the tracked target from different sensors ij And form a distance matrix D, and then calculate the distance threshold value r of the tracking target track data of different radars ij , and each element d in the distance matrix ij Compare with each tracking threshold, and form the judgment results into a confirmation matrix;

[0037] Data fusion is performed on the successfully associated tracks: the fusion center first directly fuses the data output by each different radar, and introduces the radial basis function in the fusion process to effectively reduce the errors that may be caused by interactive covariance. In the training process, some residual errors (ε m (k), m=1,..,n) and the error between the state estimate and the predicted value (X m (k)-ΦX m (k-1),m=1,...,n).

[0038] The advantages of the present invention are:

[0039] The present invention proposes a multi-radar networking method based on neural network compensation at low altitude in cities. The method takes into account the obstruction of urban trees and buildings, divides the deployment of radar detection subsystems into two aspects: airspace and ground, and then fuses the tracks of each radar based on the neural network compensation method to form the global UAV status position and situation information. Through this high-low layer combination method, different radars are deployed on demand according to the actual terrain, which not only saves costs but also realizes blind-spot detection of low-altitude UAVs in cities, providing accurate and stable UAV position and situation information for urban UAV detection systems.

[0040] The invention is based on the premise that the low-altitude background terrain in the city is complex and there are many obstructions, and the radar system is deployed on the principle of cost saving to achieve full surveillance area coverage. First, correlation is performed based on a simple probabilistic data interconnection method, and finally the correlated data is fused based on a neural network compensation method to improve accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a multi-radar networking method based on neural network compensation in urban low altitude in an embodiment of the present invention.

[0042] Figure 2 It is a track fusion graph based on neural network compensation in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] In the embodiment of the present invention, the following four radars are selected to detect low-altitude UAVs in cities. Table 1 below shows the performance indicators and costs of several radars when detecting small UAVs with an RCS of 0.01m2:

[0045] Table 1 Radar performance indicators and costs

[0046]

[0047] In order to avoid co-frequency interference with the 77G vehicle-mounted radar, the ultra-short-range frequency-modulated continuous wave radar of the present invention selects a radar with a center frequency of 24G.

[0048] Figure 1 A multi-radar networking method based on neural network compensation is proposed for low-altitude urban areas. Figure 1 As shown, the urban low-altitude multi-radar networking method based on neural network supplementation of the present invention includes the following steps:

[0049] Step 1: Select the radar performance indicators and costs as shown in Table 1 above.

[0050] Step 2: Deployment of airspace detection radars. Based on the scope of the area to be controlled, in order to save costs and ensure that the monitored area can be fully covered by the radar detection power and have a certain redundant coverage area, a mathematical optimization model is established to calculate the number and location of low-altitude surveillance radars that need to be deployed at high altitudes in the city.

[0051] Wherein, step 2 includes the following steps 21 to 23.

[0052] Step 21: According to the terrain of the area to be monitored, the approximate scope of the area is obtained.

[0053] Step 22: Establish an optimization mathematical model with minimizing cost and maximizing radar coverage area as objective functions, the number of radars and radar deployment locations as variables, and the total radar budget and surveillance space as constraints; then calculate the number and locations of radars to be deployed through a genetic optimization algorithm.

[0054] Step 23: The radar deployment position obtained by the optimization method may be anywhere in the surveillance area. However, considering that the low-altitude surveillance radar is relatively large in size and weight, has certain electromagnetic radiation, and has limited detection effect in strong ground clutter scenes, it is necessary to find a relatively high building or tower nearby according to the radar position and deploy the radar on the roof of the building or on the tower.

[0055] Step 3: Deploy ground detection radars. Deploy radars for corresponding scenarios based on the city’s terrain characteristics.

[0056] Wherein, step 3 includes the following steps 31 to 34.

[0057] Step 31: Introduce the geographical environment feature database of the monitored area, study the terrain of the monitored area, and calculate the radar deployment area, which is mainly divided into urban roads, between buildings and urban open areas.

[0058] Step 32: For urban roads, low-cost frequency modulated continuous wave radars are deployed based on urban basic transportation construction facilities. At present, there is generally a traffic crossbar every few hundred meters on urban roads, which is basically consistent with the detection distance of frequency modulated continuous wave radars. Therefore, installing the radar on the crossbar to illuminate the sky can just form a continuous ultra-low altitude coverage network. In addition, the radar system is small in size and light in weight, and will not cause extra burden on the crossbar.

[0059] Step 33: For buildings, deploy radars with different detection powers according to the distance between buildings. The radar can be deployed at 1 / 3 of the floor to illuminate the air with different illumination areas. Considering the cost, it is best that the illumination areas between adjacent buildings overlap.

[0060] During the specific implementation, ultra-short-range frequency-modulated continuous wave radar can be selected between buildings at close distances, and short-range surveillance radar can be selected between buildings at slightly longer distances.

[0061] Step 34: For open areas in the city, based on the area of ​​the area, deploy radars with detection power similar to the area of ​​the area.

[0062] Step 4: Convert and process the coordinates of radar data from different systems and unify them into the WGS-84 coordinate system.

[0063] Wherein, step 4 includes the following steps 41 to 43.

[0064] Step 41: A distributed data processing architecture is adopted, and each radar has built-in signal processing and data processing modules, which can generate track data independently.

[0065] Step 42: Use a high-precision longitude and latitude coordinate locator to measure the precise longitude and latitude information of each radar site, and then call the proj library for geographic coordinate system conversion and map projection to convert the radar track data to the WGS-84 coordinate system.

[0066] Step 43: Use the city traffic public network to transmit the converted radar track data to the data center.

[0067] Step 5: Multi-sensor data fusion processing. Considering that different radar detection areas overlap, it is necessary to match and fuse the same target from different radars into one track based on the UAV's motion position, flight speed, and flight altitude. The flow chart is as follows: Figure 2 shown.

[0068] Wherein, step 5 includes the following steps 51 and 52.

[0069] Step 51, associate the same target data from different sensors. Considering the large number of targets in the area, the calculation of the joint probability data association algorithm (JPDA) is very large and will result in combinatorial explosion. Therefore, a simple joint probability data association (CJPDA) is introduced on this basis. This interconnection method calculates the Mahalanobis distance d between the track data of the tracking targets from different sensors. ij And form a distance matrix D, and then calculate the distance threshold value r of the tracking target track data of different sensors ij , and each element d in the distance matrix ij The judgment results are compared with each tracking threshold and formed into a confirmation matrix, thereby forming a simple confirmation matrix to reduce the amount of data interconnection calculation.

[0070] Step 52: Data fusion is performed on the successfully associated tracks. In the fusion center, a relatively simple fusion is first performed on the data output by each different sensor, and then a radial basis function (RBF) is introduced in the fusion process to effectively reduce the error that may be caused by interactive covariance. In the training process, some residual errors (ε m (k), m=1,..,n) and the error between the state estimate and the predicted value (X m (k)-ΦX m (k-1),m=1,...,n).

[0071] Step 6: Conduct situation analysis on the UAV based on its motion status and characteristics.

[0072] Wherein, step 6 includes the following steps 61 and 62.

[0073] Step 61: Process the drone data information in the data center, estimate the drone's operating status, location information, size, target type, and friend-or-foe attribute information during flight, and evaluate its potential threat level based on this information.

[0074] Step 62: Send the threat level information to the integrated display and control platform, and relevant personnel will choose control methods such as expulsion, forced landing, interception or attack based on the drone situation.

[0075] The present invention provides a multi-radar networking method based on neural network compensation at low altitude in cities. The method first deploys airspace and ground radars separately according to the urban terrain on the principle of cost saving; then the radar data of different systems are processed and converted, unified to the WGS-84 coordinate system, and then the different sensor data from the same target are fused based on neural network compensation to eliminate redundant track data; finally, the situation analysis of the drone is performed based on the motion state and characteristics of the drone to provide a basis for the management and control of relevant personnel. In order to solve the problem of how to deploy radar detection subsystems under the premise of cost saving in the context of urban environment, the present invention proposes a multi-radar networking method based on neural network compensation at low altitude in cities. Through this high-low hierarchical combination method, the advantages of radars of different systems are complemented, which not only saves costs, but also realizes the detection and positioning of urban low-altitude drones without blind spots, provides accurate and stable drone position and situation information for safety management, and provides effective protection for important places and urban safety.

[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-radar networking method based on neural network compensation at low altitude in cities, characterized in that: The following steps are involved: The urban low altitude is divided into two detection spaces, airspace and ground. Multiple surveillance radars of different systems are deployed in the two detection spaces, and the radar data of different systems are unified in the coordinate system. Then, based on the neural network compensation method, the radar tracks of the same target from different radars are fused to form the global UAV status position and situation information.

2. The multi-radar networking method based on neural network compensation in urban low altitude according to claim 1 is characterized in that: The radar deployment method of the airspace is as follows: According to the terrain of the area to be monitored, the approximate scope of the area is obtained; An optimization mathematical model is established, which takes minimizing cost and maximizing radar coverage area as the objective function, the number of radars and radar deployment locations as variables, and the total radar budget and surveillance space as constraints; then the number and locations of radars to be deployed are calculated using a genetic optimization algorithm; According to the calculated radar position, find a relatively tall building or tower nearby and deploy the radar on the roof of the building or tower.

3. The multi-radar networking method based on neural network compensation in urban low altitude according to claim 1 is characterized in that: The ground radar deployment method is as follows: Introduce the geographical environment feature database of the monitored area, study the terrain of the monitored area, and calculate the radar deployment area, which is mainly divided into urban road sections, between buildings and urban open areas; For urban roads, low-cost frequency-modulated continuous wave radars are deployed based on the city's basic transportation construction facilities; for buildings, radars with different detection powers are deployed according to the distance between the buildings; for open areas in the city, radars with detection power similar to the area of ​​the area are deployed based on the area.

4. The urban low-altitude multi-radar networking method based on neural network compensation according to any one of claims 1 to 3 is characterized in that: The method to unify the coordinate system of radar data of different systems is: A distributed data processing architecture is adopted, and signal processing and data processing modules are embedded in each radar to generate track data independently. The precise longitude and latitude information of each radar site is measured, and then the proj library used for geographic coordinate system conversion and map projection is called to convert the radar track data to the WGS-84 coordinate system. The converted radar track data is transmitted to the data center using the urban traffic public network.

5. The multi-radar networking method based on neural network compensation at low altitude in cities according to any one of claims 1 to 3, characterized in that: The method for fusing different radar track data of the same target is as follows: Correlate the same target data from different radars: Use a simple joint probability data association method to calculate the Mahalanobis distance d between the track data of the tracked target from different sensors ij And form a distance matrix D, and then calculate the distance threshold value r of the tracking target track data of different radars ij , and each element d in the distance matrix ij Compare with each tracking threshold, and form the judgment results into a confirmation matrix; Data fusion is performed on the successfully associated tracks: the fusion center first directly fuses the data output by each different radar, and introduces the radial basis function in the fusion process to effectively reduce the errors that may be caused by interactive covariance. In the training process, some residual errors (ε m (k), m=1,..,n) and the error between the state estimate and the predicted value (X m (k)-ΦX m (k-1),m=1,...,n).

6. The multi-radar networking system based on neural network compensation at low altitude in the city is characterized by: include: Airspace and ground radar monitoring module: divides the urban low altitude into two detection spaces: airspace and ground, and deploys multiple surveillance radars of different systems in the two detection spaces; Data management module: unify the coordinate system of radar data of different systems; Data fusion module: Based on the neural network compensation method, the radar tracks of the same target from different radars are fused to form the global UAV status position and situation information.

7. The urban low-altitude multi-radar networking system based on neural network compensation according to claim 6 is characterized in that: The radar deployment method of the airspace is as follows: According to the terrain of the area to be monitored, the approximate scope of the area is obtained; An optimization mathematical model is established, which takes minimizing cost and maximizing radar coverage area as the objective function, the number of radars and radar deployment locations as variables, and the total radar budget and surveillance space as constraints; then the number and locations of radars to be deployed are calculated using a genetic optimization algorithm; According to the calculated radar position, find a relatively tall building or tower nearby and deploy the radar on the roof of the building or tower.

8. The urban low-altitude multi-radar networking system based on neural network compensation according to claim 6 is characterized in that: The ground radar deployment method is as follows: Introduce the geographical environment feature database of the monitored area, study the terrain of the monitored area, and calculate the radar deployment area, which is mainly divided into urban road sections, between buildings and urban open areas; For urban roads, low-cost frequency-modulated continuous wave radars are deployed based on the city's basic transportation construction facilities; for buildings, radars with different detection powers are deployed according to the distance between the buildings; for open areas in the city, radars with detection power similar to the area of ​​the area are deployed based on the area.

9. The urban low-altitude multi-radar networking system based on neural network compensation according to any one of claims 6 to 8, characterized in that: The method to unify the coordinate system of radar data of different systems is: A distributed data processing architecture is adopted, and signal processing and data processing modules are embedded in each radar to generate track data independently. The precise longitude and latitude information of each radar site is measured, and then the proj library used for geographic coordinate system conversion and map projection is called to convert the radar track data to the WGS-84 coordinate system. The converted radar track data is transmitted to the data center using the urban traffic public network.

10. The urban low-altitude multi-radar networking system based on neural network compensation according to any one of claims 6 to 8, characterized in that: The method for fusing different radar track data of the same target is as follows: Correlate the same target data from different radars: Use a simple joint probability data association method to calculate the Mahalanobis distance d between the track data of the tracked target from different sensors ij And form a distance matrix D, and then calculate the distance threshold value r of the tracking target track data of different radars ij , and each element d in the distance matrix ij Compare with each tracking threshold, and form the judgment results into a confirmation matrix; Data fusion is performed on the successfully associated tracks: the fusion center first directly fuses the data output by each different radar, and introduces the radial basis function in the fusion process to effectively reduce the errors that may be caused by interactive covariance. In the training process, some residual errors (ε m (k), m=1,..,n) and the error between the state estimate and the predicted value (X m (k)-ΦX m (k-1),m=1,...,n).

Images