Intelligent reflector assisted computational imaging method
By employing a computational imaging method assisted by an intelligent reflective surface, and utilizing the design of the reflectance coefficient and the optimal linear unbiased estimator, the problems of poor perception accuracy and imaging effect in complex environments are solved, and efficient image restoration is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2023-02-17
- Publication Date
- 2026-06-05
AI Technical Summary
In complex environments, existing wireless computational imaging methods suffer from low sensing accuracy and poor imaging results.
The computational imaging method using intelligent reflector-assisted reflectivity involves designing a reflectivity vector θopt, utilizing a base station and intelligent reflector in conjunction with a terminal device to transmit sensing signals, and combining this with an optimal linear unbiased estimator to recover the object image and optimize the reflectivity to reduce mean square error.
It improves perception accuracy and imaging quality in complex environments, simplifies the mean square error expression, is applicable to various scenarios, and has low computational complexity.
Smart Images
Figure CN116451417B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wireless communication, and more particularly to a smart reflective surface-assisted computational imaging method in wireless networks. Background Technology
[0002] In recent years, smart reflector technology has attracted widespread attention in the field of wireless communication due to its low manufacturing cost, flexible deployment location, and convenient controllability. Smart reflectors are composed of reconfigurable metamaterial elements and reflect incident signals by introducing a predetermined phase shift. This phase shift is controlled by the base station through a backhaul control link based on the received external signal and can be dynamically adjusted in real time. When the communication link with the target within a certain range is weak or blocked, smart reflectors can be deployed to enhance communication. Currently, smart reflectors are being extensively studied, including for extending the range of users with blocked direct links, physical layer security, and unmanned aerial vehicle (UAV) communication.
[0003] With the advent of the 6G era, wireless sensing applications utilizing the widespread availability of wireless signals have attracted increasing research interest, and computational imaging is one of the best wireless sensing technologies. Unlike methods based on wearable devices or surveillance cameras, wireless computational imaging does not require direct contact with the target. Its basic principle is that the influence of the target object on the propagation of wireless signals can be recognized by the receiver; that is, by utilizing the scattering rate of the target object itself, the image of the object can be reconstructed through calculation. Wireless computational imaging technology can be widely applied to many scenarios in daily life, such as surveillance, autonomous driving, environmental assistance, and virtual reality. In these applications, the environment, especially the urban environment, is complex and changeable with numerous obstacles. Improving sensing accuracy is crucial. By utilizing intelligent reflective surface technology, the wireless sensing path can be increased or enhanced, just like communication methods, thereby achieving rapid and accurate imaging.
[0004] Therefore, combining intelligent reflective surface technology and computational imaging technology can overcome environmental limitations, improve perception accuracy, and achieve efficient imaging. Summary of the Invention
[0005] To improve the perception effect and imaging accuracy in complex environments, this invention proposes a computational imaging method assisted by an intelligent reflective surface.
[0006] The specific technical solution adopted in this invention is as follows:
[0007] This invention provides a computational imaging method assisted by a smart reflective surface, comprising the following steps:
[0008] 1) The system consists of K single-antenna terminal devices, a passive intelligent reflector containing L reflective elements, and a base station with M antennas. The controller of the intelligent reflector is connected to the base station, and a three-dimensional object to be imaged exists simultaneously. The reflection coefficient vector of the L reflective elements is... The reflection coefficient of the l-th reflecting unit is ρ l It's the amplitude. It is phase;
[0009] 2) Discretize the environmental space containing the 3D object to be imaged to obtain N point bodies of the same size. The scattering coefficient vector of these N point bodies is x = [x1, ..., x2]. N ] T The scattering coefficient of the nth point is x. n Its value is between 0 and 1;
[0010] 3) Obtain channel state information from the terminal device to the 3D object to be imaged, from the 3D object to the smart reflector, from the smart reflector to the base station, and from the 3D object to the base station. Based on the obtained channel state information, use the mean square error of the scattering coefficient vector of the 3D object estimated by the base station as the objective function. Using a reflection coefficient design method for a smart reflector, obtain the reflection coefficient θ that minimizes the imaging mean square error. opt The base station is based on θ opt Configure intelligent reflective surface;
[0011] 4) All terminal devices simultaneously transmit sensing signals to the 3D object to be imaged in the environment, and the signal vector composed of all sensing signals is: Where p k and s k These are the transmission power and transmission signal of the kth terminal device, respectively. After the signal is scattered by the three-dimensional object to be imaged, part of it reaches the base station directly, and part of it reaches the base station after being reflected by the intelligent reflective surface.
[0012] 5) After receiving the signal, the base station uses the best linear unbiased estimator to obtain the best estimated vector of the three-dimensional object to be imaged. Covariance Matrix Then, the image of the three-dimensional object to be imaged is recovered.
[0013] As a preferred embodiment, the method for designing the reflection coefficient of a smart reflective surface in step 3) is as follows:
[0014] a) Let the channel matrices from the terminal device to the 3D object to be imaged, from the 3D object to the smart reflector, from the smart reflector to the base station, and from the 3D object to the base station be respectively... and Where α1, α2, α3, and β are large-scale fading coefficients based on path loss and shadow fading, respectively; H TU、 H BR H RT and H BT These are small-scale fading coefficient matrices, both of which follow a standard normal distribution;
[0015] b) Initialize the reflection coefficient vector of the smart reflector as follows: The amplitude is Randomly generated phase
[0016] c) The signal received by the base station is
[0017]
[0018] Where diag(·) denotes diagonalization, and n is a matrix with a mean of 0 and a covariance matrix of R = σ. 2 I M Additive white Gaussian noise; according to diag(x)H TU s = diag(H TU s)x, transforming the expression into a standard linear model:
[0019]
[0020] in
[0021] d) The scattering coefficient of the three-dimensional object to be imaged is obtained by the base station through optimal linear unbiased estimation. Its covariance matrix is The scattering coefficient of the nth point object estimated by the base station Compared with the actual scattering coefficient x n The mean square error is
[0022]
[0023] in Unbiased estimation but Scattering coefficient vector to be estimated The mean square error is in(·) H This represents the conjugate transpose of a matrix, (·). -1 This represents finding the inverse of a matrix. The expression `[·]` represents the expected value, and `var(·)` represents the variance. nn tr[·] represents the element in the nth row and nth column of the matrix, and tr[·] represents finding the trace of the matrix;
[0024] e) Based on the acquired channel state information from the terminal device to the 3D object to be imaged, from the 3D object to the smart reflector, from the smart reflector to the base station, and from the 3D object to the base station, the transmit power and transmit signal are fixed. Using the mean square error of the 3D object to be imaged as the objective function, in |θ l Given |=1, l=1,...,L, the mathematical model for minimizing the mean square error by optimizing variable θ is expressed as min θ MSE;
[0025] f) Let Λ = G + Adiag(θ)B, where B = H RT diag(H TU s), then
[0026] MSE = σ 2 tr[((Adiag(θ)B) H (Adiag(θ)B)) -1 ]
[0027] Again Where A = [a1, ..., a2] L ], B = [b1, ..., b L ] T a i , Let represent the i-th column of A and B respectively, (·) T This indicates transpose, transforming the mean squared error into...
[0028]
[0029] in It is a non-singular matrix, invertible, and its inverse is... And its rank (D) l If ) = 1, then the mathematical problem model in step e) is transformed into
[0030] g) Fixing other reflection coefficients, calculate the reflection coefficient of the l-th reflection unit.
[0031] Where I N For an N-dimensional identity matrix; Perform Schur decomposition to obtain in It is a unitary matrix. Let be an upper triangular matrix, and its eigenvalues and sums be... Same, and ∑ l It can be represented as and but make in Lemma of Inverse of Block Matrix Again in but From |θ l |=1 indicates that Substitute f l get
[0032]
[0033] Take its partial derivative and let get
[0034]
[0035] Where d = ε H w, c = e H e, b = tr(Wee H Similarly, perform the same calculation on the reflection coefficients of other reflecting units, iterating until convergence, to obtain the optimal reflection coefficient vector of the intelligent reflecting surface.
[0036]
[0037] h) Base station according to θ opt Configure the reflectivity of the intelligent reflective surface.
[0038] Preferably, step 5) is as follows:
[0039] The base station uses an optimal linear unbiased estimator based on the received sensing signals to obtain the optimal scattering coefficient vector of the three-dimensional object to be imaged. And recover the image of the three-dimensional object to be imaged, where
[0040] The beneficial effects of this invention are as follows: The intelligent reflective surface-assisted computational imaging method proposed in this invention solves the problems of low perception accuracy and poor computational imaging effect in complex environments. The reflection coefficient design method of the intelligent reflective surface proposed in this invention simplifies the mean square error expression, has the characteristics of low computational complexity, and is applicable to any scenario. Attached Figure Description
[0041] Figure 1 This is a block diagram of a computational imaging method assisted by a smart reflective surface;
[0042] Figure 2 This is a comparison of the imaging images of the intelligent reflective surface under four different configurations;
[0043] Figure 3 It compares the mean square error of the object to be imaged under different numbers of reflective elements. Detailed Implementation
[0044] The present invention will be further described and illustrated below with reference to the accompanying drawings and specific embodiments.
[0045] In this embodiment, the intelligent reflective surface-assisted computational imaging method is as follows: Figure 1 As shown, considering only the uplink, the reflection coefficient of the intelligent reflector is initialized. The base station adjusts the reflection coefficient of the intelligent reflector to minimize the mean square error of the object to be imaged. All multiple single-antenna user equipment simultaneously transmit sensing signals with the same power to the base station in the environment. After being scattered by an object to be imaged, and then reflected by an intelligent reflector containing L reflection elements, the signal reaches the base station. The base station uses the best linear unbiased estimator to achieve imaging of the object to be imaged.
[0046] The intelligent reflective surface-assisted computational imaging method used in this embodiment specifically includes the following steps:
[0047] 1) The system consists of K single-antenna terminal devices and a passive intelligent reflector containing L reflector elements, the reflection coefficient vector of which is: The reflection coefficient of the l-th reflecting unit is ρ l It's the amplitude. It consists of a phase and a base station with M antennas, wherein the controller of the intelligent reflector is connected to the base station, and there is a three-dimensional object to be imaged.
[0048] 2) Discretize the environmental space containing the three-dimensional object to obtain N point bodies of the same size. The scattering coefficient vector of these N point bodies is x = [x1, ..., x2]. N ] T The scattering coefficient of the nth point is x. n Its value is between 0 and 1;
[0049] 3) Obtain channel state information from the terminal device to the object to be imaged, from the object to the smart reflector, from the smart reflector to the base station, and from the object to the base station. Based on the channel state information, use the mean square error of the scattering coefficient vector of the object to be imaged estimated by the base station as the objective function. Using a reflection coefficient design method for a smart reflector, obtain the reflection coefficient θ that minimizes the imaging mean square error. opt The base station is based on θ optConfigure intelligent reflective surface;
[0050] 4) All terminal devices simultaneously transmit sensing signals to the object to be imaged in the environment, and the signal vector composed of all sensing signals is: Where p k and s k These are the transmission power and transmission signal of the kth terminal device, respectively. After the signal is scattered by the object to be imaged, part of it reaches the base station directly, and part of it reaches the base station after being reflected by the intelligent reflective surface.
[0051] 5) After receiving the signal, the base station uses the best linear unbiased estimator to obtain the best estimated vector of the object to be imaged. Covariance Matrix Then, the image of the three-dimensional object to be imaged is recovered.
[0052] In this embodiment, the specific implementation steps of the intelligent reflective surface reflection coefficient design method described in step 3) above are as follows:
[0053] a) Let the channel matrices from the terminal device to the object to be imaged, from the object to the smart reflector, from the smart reflector to the base station, and from the object to the base station be respectively... and Where α1, α2, α3, and β are large-scale fading coefficients based on path loss and shadow fading, respectively, and H TU H BR H RT and H BT These are small-scale fading coefficient matrices, both of which follow a standard normal distribution;
[0054] b) Initialize the reflection coefficient vector of the smart reflector as follows: The amplitude is Randomly generated phase
[0055] c) The signal received by the base station is
[0056]
[0057] Where diag(·) denotes diagonalization, and n is a matrix with a mean of 0 and a covariance matrix of R = σ. 2 I M Additive white Gaussian noise, according to diag(x)H TU s = diag(H TU s)x, transforming the expression into a standard linear model,
[0058]
[0059] in
[0060] d) The scattering coefficient of the object to be imaged is obtained by the base station through the best linear unbiased estimation. Its covariance matrix is The scattering coefficient of the nth point object estimated by the base station Compared with the actual scattering coefficient x n mean square error
[0061]
[0062] in Unbiased estimation but Scattering coefficient vector to be estimated The mean square error is in(·) H This represents the conjugate transpose of a matrix, (·). -1 This represents finding the inverse of a matrix. [·] indicates the expected value, Var(·) indicates the variance, and [·] nn tr[·] represents the element in the nth row and nth column of the matrix, and tr[·] represents finding the trace of the matrix;
[0063] e) Based on the acquired channel state information from the terminal device to the object to be imaged, from the object to the smart reflector, from the smart reflector to the base station, and from the object to the base station, the transmit power and transmit signal are fixed, and the mean square error of the object to be imaged is used as the objective function, in |θ l Given |=1, l=1,...,L, the mathematical model for minimizing the mean square error by optimizing variable θ is expressed as min θ MSE;
[0064] f) Let Λ = G + Adiag(θ)B, where B = H RT diag(H TU s), then
[0065] MsE = σ 2 tr[((Adiag(θ)B) H (Adiag(θ)B)) -1 ]
[0066] Again Where A = [a1, ..., a2] L ], B = [b1, ..., b L ] T a i , Let represent the i-th column of A and B respectively, (·) T This indicates transpose, transforming the mean squared error into...
[0067]
[0068] in It is a non-singular matrix, invertible, and its inverse is... And its rank (D) l If ) = 1, then the mathematical problem model in e) is transformed into
[0069] g) Fixing other reflection coefficients, calculate the reflection coefficient of the l-th reflection unit.
[0070] Among them I N For an N-dimensional identity matrix; Perform Schur decomposition to obtain in It is a unitary matrix. Let be an upper triangular matrix, and its eigenvalues and sums be... Same, and ∑ l It can be represented as and but make in Lemma of Inverse of Block Matrix Again in but From |θ l |=1 indicates that Substitute f l get
[0071]
[0072] Take its partial derivative and let get
[0073]
[0074] Where d = ε H w, c = e H e, b = tr(Wee H The same calculation is performed on the reflection coefficients of other reflecting units, iterating until convergence, to obtain the optimal reflection coefficient vector of the intelligent reflecting surface.
[0075] h) Base station according to θopt Configure the reflection coefficient of the intelligent reflective surface;
[0076] In this embodiment, the specific implementation steps of the optimal linear unbiased estimation imaging method described in step 5) above are as follows:
[0077] The base station uses an optimal linear unbiased estimator based on the received sensing signals to obtain the optimal scattering coefficient vector of the object to be imaged. And recover the image of the object, where
[0078] Computer simulations show that, Figure 2 As shown, in the computational imaging method with intelligent reflective surface assistance proposed in this invention, compared with the original target image, the imaging effect without intelligent reflective surface assistance is poor, the imaging effect with fixed intelligent reflective surface reflectivity is better, and the imaging effect obtained by optimizing the reflectivity of intelligent reflective surface through the proposed optimization method is the best. Figure 3 This indicates that in the method proposed in this invention, as the number of reflective units on the intelligent reflective surface increases, the signal transmission path increases, the object perception accuracy becomes higher, and the mean square error of the imaging becomes smaller. Therefore, this invention provides an effective computational imaging method for wireless networks based on intelligent reflective surfaces.
[0079] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.
Claims
1. A computational imaging method assisted by an intelligent reflective surface, characterized in that, Includes the following steps: 1) The system consists of K A single-antenna terminal device, a device containing L A passive intelligent reflector with one reflector element and one antenna with [number of elements]. M The system consists of a base station, with the controller of the intelligent reflective surface connected to the base station, and a three-dimensional object to be imaged is present simultaneously; wherein, L The reflection coefficient vector of each reflecting unit is , No. l The reflection coefficient of each reflecting unit is , It's the amplitude. It is phase; 2) Discretize the environmental space containing the 3D object to be imaged to obtain... N A dot-like object of the same size, N The scattering coefficient vector of the point-like body is Among them, the first n The scattering coefficient of the point-like object is Its value is between 0 and 1; 3) Obtain channel state information from the terminal device to the 3D object to be imaged, from the 3D object to the smart reflector, from the smart reflector to the base station, and from the 3D object to the base station. Based on the obtained channel state information, use the mean square error of the scattering coefficient vector of the 3D object estimated by the base station as the objective function. Utilize a reflection coefficient design method for a smart reflector to obtain the reflection coefficient that minimizes the imaging mean square error. The base station according to Configure intelligent reflective surface; 4) All terminal devices simultaneously transmit sensing signals to the 3D object to be imaged in the environment. The signal vector composed of all sensing signals is: ;in and These are the transmission power and transmission signal of the kth terminal device, respectively. After the signal is scattered by the three-dimensional object to be imaged, part of it reaches the base station directly, and part of it reaches the base station after being reflected by the intelligent reflective surface. 5) After receiving the signal, the base station uses the best linear unbiased estimator to obtain the best estimated vector of the three-dimensional object to be imaged. Covariance Matrix Then, the image of the three-dimensional object to be imaged is recovered; The specific method for designing the reflection coefficient of a smart reflective surface in step 3) is as follows: a) Let the channel matrices from the terminal device to the 3D object to be imaged, from the 3D object to the smart reflector, from the smart reflector to the base station, and from the 3D object to the base station be respectively... , , and ;in , , and These are the large-scale fading coefficients based on path loss and shadow fading, respectively. , , and These are small-scale fading coefficient matrices, both of which follow a standard normal distribution; b) Initialize the reflection coefficient vector of the smart reflector as follows: The amplitude is Randomly generated phase ; c) The signal received by the base station is ;in Indicates diagonalization, It has a mean of 0 and a covariance matrix of Additive white Gaussian noise; according to Transform the expression into a standard linear model: ;in ; d) The scattering coefficient of the three-dimensional object to be imaged is obtained by the base station through optimal linear unbiased estimation. Its covariance matrix is The base station estimated the first n Scattering coefficient of a point-like object Compared with the actual scattering coefficient The mean square error is ； in Unbiased estimation ,but ; Scattering coefficient vector to be estimated The mean square error is ,in This represents the conjugate transpose of a matrix. This represents finding the inverse of a matrix. Indicates the expectation. This indicates the calculation of variance. Represents the matrix of the first n Line number m Column elements, Represents finding the trace of a matrix; e) Based on the acquired channel state information from the terminal device to the 3D object to be imaged, from the 3D object to the smart reflector, from the smart reflector to the base station, and from the 3D object to the base station, the transmit power and transmit signal are fixed, and the mean square error of the 3D object to be imaged is used as the objective function. Under the condition of optimizing variables To minimize the mean square error, the mathematical model is expressed as follows: ; f) Let ,in , , ,but ; then order ,in , , Represent A , B The i List, This indicates transpose, transforming the mean squared error into... ; in is a non-singular matrix, invertible, and its inverse is . , And its rank Then the mathematical problem model in step e) is transformed into ; g) Fixing other reflection coefficients, calculate the first... l Reflection coefficient of each reflecting unit ; , in for N Dimensional unit matrix; for Perform Schur decomposition to obtain , ,in It is a unitary matrix. Let be an upper triangular matrix, and its eigenvalues and sums be... Same, and , It can be represented as ,and ,but ;make ,in According to the lemma of finding the inverse of a block matrix, , and then ,in , ,but ,Depend on It can be known Substitute get Take the partial derivative with respect to it. make ,get ;in , , The same calculation is performed on the reflection coefficients of other reflecting units, iterating until convergence, to obtain the optimal reflection coefficient vector of the intelligent reflecting surface. ; h) Base station according to Configure the reflectivity of the intelligent reflective surface.
2. The computational imaging method assisted by an intelligent reflective surface according to claim 1, characterized in that, Step 5) is as follows: The base station uses an optimal linear unbiased estimator based on the received sensing signal to obtain the optimal scattering coefficient vector of the three-dimensional object to be imaged. And recover the image of the three-dimensional object to be imaged, where .