Method for identifying and locating tiny objects in complex environments using SISO antennas

Abstract

The present invention provides a method for identifying and locating tiny objects in a complex environment by using a SISO antenna, comprising the following steps: S1. establishing a cubic vacuum resonant cavity and placing a partition; S2. setting two ports S3 inside the cubic vacuum resonant cavity; feeding port one to send a signal, while port two is used to receive a signal; S4. exporting the received signal of port two, performing a time reversal operation, and doubling the time-reversed signal; S5. placing a metal block in the cubic vacuum resonant cavity, using the processed signal to excite port two, while port one is used to receive the signal sent by port two; S6. obtaining a time-reversed phase ripple image; S7. obtaining time-frequency images of objects of different shapes and at different positions; S8. training a cascade neural network consisting of a classification network and a positioning network; and S9. testing the accuracy of the neural network to evaluate the robustness of the neural network.

Description

Technical Field

[0001] The present invention relates to the field of wireless signal monitoring and neural network technology, and in particular to a method for identifying and locating tiny objects in a complex environment using a SISO antenna. Background Art

[0002] The perception of non-cooperative targets plays an important role in security, counter-terrorism, health monitoring, and other fields. Methods for achieving non-cooperative target perception include video surveillance, infrared monitoring, radio frequency identification (RFID) monitoring, and wireless signal-based monitoring. Among them, wireless signal-based monitoring methods can perceive non-cooperative targets in complex environments by analyzing changes in radio signal propagation characteristics, and have anti-multipath performance that other methods do not have. However, traditional wireless signal-based monitoring methods have difficulty perceiving subwavelength (especially deep subwavelength) non-cooperative targets. Therefore, there is an urgent need for a method to perceive subwavelength (especially deep subwavelength) non-cooperative targets and simultaneously solve the fitting or regression problem of large amounts of data. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention provides a method for identifying and locating tiny objects in complex environments using SISO antennas, which can perceive subwavelength (especially deep subwavelength) non-cooperative targets and simultaneously solve the fitting or regression problems of large amounts of data.

[0004] The above technical objectives of the present invention are achieved through the following technical solutions:

[0005] A method for identifying and locating tiny objects in a complex environment using a SISO antenna includes the following steps:

[0006] S1. Create a cubic vacuum resonant cavity and place a partition at the top, offset from the center, of the cavity.

[0007] S2. Two ports are provided inside the cubic vacuum resonant cavity, namely port 1 and port 2, and port 1 and port 2 are located on both sides of the partition respectively;

[0008] S3. Set each port type to the S-Parameter type in the discrete port. Port 1 is used to feed the signal, while port 2 is used to receive the signal. After 100 ns of simulation, the received signal of port 2 is obtained.

[0009] S4. The received signal of port 2 is derived, a time reversal operation is performed, and the time-reversed signal is doubled, the extended signal value is set to 0, and the processed signal is obtained;

[0010] S5. Place a metal block inside the cubic vacuum resonator and use the processed signal to excite port 2. Port 1 then receives the signal sent from port 2. After 200 ns of simulation, the received signal at port 1 is obtained.

[0011] S6. Derive and calculate the received signal of port 1 to obtain a time-reversed phase ripple image;

[0012] S7. Change the shape and placement of the metal block and repeat steps S5 and S6 to obtain time-frequency images of objects of different shapes and positions;

[0013] S8. Using a portion of the time-frequency images in S7, train a cascade neural network consisting of a classification network and a localization network;

[0014] S9. Use the time-frequency image from another part of S7 to test the accuracy of the neural network to evaluate the robustness of the neural network.

[0015] The present invention is further configured such that: in step S1, the placed partition is an ideal partition with no thickness.

[0016] The present invention is further configured as follows: in step S4, a time reversal operation is performed by keeping the time order unchanged and reversing the signal.

[0017] The present invention has the following beneficial effects:

[0018] The present invention can obtain time-frequency images of different objects and different positions under complex conditions in a cubic vacuum resonant cavity, and realizes the function of accurately identifying different objects and accurately locating different positions using neural network means. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The figure is a schematic diagram of the overall process;

[0020] Figure 2 Schematic diagram of the cubic vacuum resonant cavity, partition and port in the example;

[0021] Figure 3 This is a schematic diagram of placing an object at a certain position in a cubic vacuum resonant cavity in an example;

[0022] Figure 4 The time-frequency image of a certain object at a certain position obtained by processing;

[0023] Figure 5 It is the structural diagram of the cascade neural network;

[0024] Figure 6 The accuracy curves of the training set and test set of the classification network in the cascade neural network;

[0025] Figure 7 This is the accuracy heat map of the classification network in the cascade neural network;

[0026] Figure 8 Accuracy curves of the training set and test set for the cylinder positioning network in the cascade neural network;

[0027] Figure 9 Heatmap of the accuracy of the cylinder localization network in the cascaded neural network. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0029] A method for identifying and locating tiny objects in a complex environment using a SISO antenna includes the following steps: first, a cubic vacuum resonant cavity is established in the CST simulation software, such as Figure 2 shown.

[0030] Among them, the side length of the cubic vacuum resonance cavity is L = 60 cm. In order to allow the signal to pass through the object in the cavity before being received by the receiving antenna, it is necessary to add an ideal non-thickness partition with a length of 60 cm and a width of b = 15 cm on the upper part of the cube. In order to eliminate the influence of the central symmetry of the non-thickness partition on the excessive similarity of the signals, the non-thickness partition is placed on the upper wall of the cubic vacuum resonance cavity, 5 cm to the left of the center, that is, a = 25 cm.

[0031] Two ports are set inside the cubic vacuum resonator, namely port 1 and port 2, and port 1 and port 2 are located on both sides of the partition respectively; take the side where the object is to be placed as the bottom, its center as the origin, the side facing the partition as the z-axis, and the left side of this as the x-axis to establish a spatial rectangular coordinate system. The two ports are placed at: port 1 is Port 2 is like Figure 2 In the example, port one is r1 and port two is r2.

[0032] Set port 1 and port 2 as S-Parameter types in discrete ports. Port 1 is fed to send signals, while port 2 is used to receive signals. After 100ns of simulation, the received signal of port 2 is obtained. The received signal of port 2 is exported and time-reversed. The signal after time-reversal is extended by half and the extended signal value is set to 0 to obtain the processed signal. A metal block is placed in the cubic vacuum resonant cavity. Here, a cylinder is used as an example. Figure 3The object size and position in the cavity are set as shown in the figure. Use the processed signal to stimulate port 2. At this time, port 1 is used to receive the signal sent by port 2. After 200ns of simulation, the received signal of port 1 is obtained. The received signal of port 1 is derived and calculated to obtain the following: Figure 4 The time-reversed phase ripple image shown in FIG. 4 is obtained; the shape and placement of the metal block are then changed, and the above steps are repeated to obtain time-frequency images of objects of different shapes and positions;

[0033] After obtaining a rich library of time-frequency images, use the training set and follow the steps below: Figure 5 The cascade network architecture shown here trains a neural network that can be used for both recognition and localization. After training the neural network, the robustness of the neural network is evaluated using a test set. The accuracy curves for the training and test sets of the two classification and cylinder localization networks in the cascade network are shown in the figure below. Figure 6 、 8 As shown, the accuracy heat map corresponds to Figure 9 As shown. Figure 6 It is the accuracy curve of the training set and test set of the classification small network, Figure 8 It is the accuracy heat map of the classification small network; Figure 7 This is the accuracy curve of the training set and test set of the cylinder positioning small network, Figure 9 This is a heatmap of the accuracy of the cylinder localization network. It can be seen that the accuracy of the classification network on both the test and training sets reaches 100%, and the accuracy of the cylinder localization network on both the training and test sets also reaches 100%.

[0034] Since the lengths of the obtained signals are not exactly the same, we process the time-frequency images into a two-dimensional matrix of 201×155 before training and testing the network.

[0035] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for identifying and locating small objects in a complex environment using a SISO antenna, characterized by: The following steps are involved: S1. Create a cubic vacuum resonant cavity and place a partition at the top, offset from the center, of the cavity. S2. Two ports are provided inside the cubic vacuum resonant cavity, namely port 1 and port 2, and port 1 and port 2 are located on both sides of the partition respectively; S3. Set each port type to the S-Parameter type in the discrete port. Port 1 is used to feed the signal, while port 2 is used to receive the signal. After 100 ns of simulation, the received signal of port 2 is obtained. S4. The received signal of port 2 is derived, a time reversal operation is performed, and the time-reversed signal is doubled, the extended signal value is set to 0, and the processed signal is obtained; S5. Place a metal block inside the cubic vacuum resonator and use the processed signal to excite port 2. Port 1 then receives the signal sent from port 2. After 200 ns of simulation, the received signal at port 1 is obtained. S6. Derive and calculate the received signal of port 1 to obtain a time-reversed phase ripple image; S7. Change the shape and placement of the metal block and repeat steps S5 and S6 to obtain time-frequency images of objects of different shapes and positions; S8. Using a portion of the time-frequency images in S7, train a cascade neural network consisting of a classification network and a localization network; S9. Use the time-frequency image from another part of S7 to test the accuracy of the neural network to evaluate the robustness of the neural network.

2. The method for identifying and locating small objects in a complex environment using a SISO antenna according to claim 1, wherein: In step S1 , the placed partition is an ideal partition without thickness.

Images