CN111065123B - Ground signal map recovery method and system - Google Patents
Ground signal map recovery method and system Download PDFInfo
- Publication number
- CN111065123B CN111065123B CN201911326197.7A CN201911326197A CN111065123B CN 111065123 B CN111065123 B CN 111065123B CN 201911326197 A CN201911326197 A CN 201911326197A CN 111065123 B CN111065123 B CN 111065123B
- Authority
- CN
- China
- Prior art keywords
- signal
- matrix
- ground
- area
- ground signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000011084 recovery Methods 0.000 title claims abstract description 57
- 238000000034 method Methods 0.000 title claims abstract description 47
- 239000011159 matrix material Substances 0.000 claims abstract description 274
- 238000003062 neural network model Methods 0.000 claims abstract description 49
- 238000013507 mapping Methods 0.000 claims description 75
- 238000005562 fading Methods 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 239000007788 liquid Substances 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 12
- 238000005070 sampling Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 7
- 238000010276 construction Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000012549 training Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000012358 sourcing Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/08—Testing, supervising or monitoring using real traffic
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a method and a system for restoring a ground signal map. The method comprises the following steps: acquiring an aerial signal matrix; the aerial signal matrix is a sparse matrix; acquiring a scene type of a signal area to be recovered; the scene types comprise a first scene type and a second scene type; when the scene type of the signal area to be recovered is a first scene type, recovering the signal of each grid of the ground area according to the aerial signal matrix based on the neural network model to obtain a recovered ground signal map; when the scene type of the signal area to be recovered is a second scene type, acquiring a ground signal matrix; non-zero elements in the ground signal matrix are ground signals of grids in the corresponding area; and recovering the signals of the grid of the area without the signal data on the ground by adopting a recovery method of accurate matrix filling according to the aerial signal matrix and the ground signal matrix to obtain a recovered ground signal map. The invention can improve the accuracy of signal map recovery.
Description
Technical Field
The invention relates to the field of signal map recovery, in particular to a method and a system for recovering a ground signal map.
Background
The signal map is constructed for key roles related to the construction of the smart city, such as spectrum detection, Location Based Service (LBS) and network state monitoring. For 5G to be deployed, a good signal map can clearly present the distribution and state of signals, and can visually find out a faulty cell and guide a decision maker to make a correct decision. In order to reduce the construction cost, most of the signal maps are constructed in a mode of crowd sensing and data filling, and data recovery is performed by collecting signals of partial areas and utilizing the space-time correlation of the signals.
The traditional signal acquisition mode needs special persons and professional equipment to acquire an area, and is time-consuming, labor-consuming and expensive. In contrast, crowd-sourcing awareness utilizes sensors in smartphones and wearable devices to encourage crowd-sourced participants to do this, greatly reducing signal acquisition costs and improving flexibility. Because the wireless signals have strong space-time correlation, only a part of data needs to be collected by using crowd sensing, and the rest data can be recovered by utilizing the space-time correlation of the wireless signals and combining a certain data filling algorithm, such as compressed sensing, a Gaussian process and matrix filling. The gaussian process is a probability-based method for recovering signals at other points in the region based on modeling the relationship between signal fluctuations and reference points. The matrix filling utilizes the low rank of the matrix to recover the missing data, and the formed wireless signal matrix has low rank because the wireless signals have strong space-time correlation, and can be used for recovering the missing wireless signal data. The method of compressed sensing is initially applied to signal processing, and two conditions need to be satisfied to use it: sparsity and irrelevancy. I.e. the signal is sparse in a certain domain and the observation matrix and the sparse representation basis of the signal are uncorrelated. For the sparse signal matrix to be recovered, the above two conditions are satisfied, so that it can be used to recover data.
However, although these methods applied to signal map construction consider the spatial correlation of signals, they only consider the spatial correlation of signals from two-dimensional space, and the correlation of signals at different heights is not considered, so the signal map construction result is not accurate.
Disclosure of Invention
The invention aims to provide a method and a system for restoring a ground signal map so as to improve the accuracy of restoring the signal map.
In order to achieve the purpose, the invention provides the following scheme:
a ground signal map recovery method, comprising:
acquiring an aerial signal matrix of a signal area to be recovered; the signal area to be recovered comprises M multiplied by N area lattices; the aerial signal matrix is a sparse matrix, and nonzero elements in the aerial signal matrix are aerial signals of corresponding region lattices;
acquiring the scene type of the signal area to be recovered; the scene types comprise a first scene type and a second scene type; the signal fluctuation range in the first scene type is smaller than a first threshold value, the signal fluctuation range in the second scene type is larger than a second threshold value, and the first threshold value is smaller than the second threshold value;
when the scene type of the signal area to be recovered is the first scene type, recovering the signal of each grid of the ground area according to an aerial signal matrix based on a neural network model to obtain a recovered ground signal map;
when the scene type of the signal area to be recovered is the second scene type, acquiring a ground signal matrix; non-zero elements in the ground signal matrix are ground signals of grids in the corresponding area;
and recovering the signals of the grid of the area without the signal data on the ground by adopting a recovery method of accurate matrix filling according to the aerial signal matrix and the ground signal matrix to obtain a recovered ground signal map.
Optionally, when the scene type of the signal area to be restored is the first scene type, restoring, based on a neural network model, a signal of each area grid on the ground according to an aerial signal matrix to obtain a restored ground signal map, where the method further includes:
inputting sample data into the neural network model; the sample data is an aerial signal matrix sample; the neural network model comprises two convolution layers, two pooling layers and a full-connection layer;
acquiring a recovered ground signal matrix output by the neural network model;
and updating the loss function of the neural network model according to the recovered ground signal matrix and the original ground signal matrix, and continuing to train the neural network model.
Optionally, the recovering method of filling an accurate matrix according to the aerial signal matrix and the ground signal matrix is used to recover the signal of the grid in the area without signal data on the ground, so as to obtain a recovered ground signal map, and the method specifically includes:
mapping the aerial signal matrix to the ground based on a channel fading model to obtain a ground signal mapping matrix; each nonzero element in the ground signal mapping matrix is a ground mapping signal of a corresponding area grid; the ground mapping signal is an average value of all aerial signals mapped to the area grid;
fusing the ground signal mapping matrix and the ground signal matrix, and determining a signal of each area grid to obtain a ground signal updating matrix;
and updating the matrix according to the ground signal, and obtaining a recovered ground signal map by adopting a recovery method of accurate matrix filling.
Optionally, the fusing the ground signal mapping matrix and the ground signal matrix to determine the signal of each area grid to obtain a ground signal update matrix specifically includes:
for each area lattice, judging whether the ground signal corresponding to the area lattice is zero or not;
when the ground signal corresponding to the area grid is not zero, determining the ground signal as a ground signal update value corresponding to the area grid;
when the ground signal corresponding to the area grid is zero, judging whether the ground mapping signal corresponding to the area grid is zero or not;
when the ground mapping signal corresponding to the area lattice is zero, determining that the ground signal updating value corresponding to the area lattice is zero;
when the ground mapping signal corresponding to the area grid is not zero, determining the ground mapping signal as a ground signal updating value corresponding to the area grid;
and determining the ground signal updating matrix according to the ground signal updating values corresponding to all the area grids.
Optionally, the updating the matrix according to the ground signal and obtaining the recovered ground signal map by using a recovery method of accurate matrix filling specifically include:
according to the formula X ═ U ∑ VTDecomposing the ground signal update matrix; wherein X is the ground signal update matrix, U is an MxM unitary matrix, Sigma is a singular value matrix, and VTIs an NxN unitary matrix;
retaining gamma singular values, dividing the retained singular values in a singular value matrixSetting the singular values except the singular value to zero to obtain a singular value updating matrix sigmaγ；
Updating the matrix according to the singular values, using formula Xγ＝UγΣγVγ TObtaining the recovered ground signal map Xγ(ii) a The recovered ground signal map XγUpdating an approximation matrix of a matrix X for the ground signal; wherein, UγA matrix consisting of the first k columns of elements of a unitary matrix of order U, Vγ TIs a unitary matrix V of orderTA matrix of the first k columns of elements.
The invention also provides a system for recovering the ground signal map, which comprises:
the aerial signal matrix acquisition module is used for acquiring an aerial signal matrix of a signal area to be recovered; the signal area to be recovered comprises M multiplied by N area lattices; the aerial signal matrix is a sparse matrix, and nonzero elements in the aerial signal matrix are aerial signals of corresponding region lattices;
a scene type obtaining module, configured to obtain a scene type of the signal region to be restored; the scene types comprise a first scene type and a second scene type; the signal fluctuation range in the first scene type is smaller than a first threshold value, the signal fluctuation range in the second scene type is larger than a second threshold value, and the first threshold value is smaller than the second threshold value;
the first recovery module is used for recovering the signal of each area grid on the ground according to an aerial signal matrix based on a neural network model when the scene type of the signal area to be recovered is the first scene type to obtain a recovered ground signal map;
a ground signal matrix obtaining module, configured to obtain a ground signal matrix when the scene type of the signal area to be recovered is the second scene type; non-zero elements in the ground signal matrix are ground signals of grids in the corresponding area;
and the second recovery module is used for recovering the signals of the grid in the area without the signal data on the ground by adopting a recovery method of accurate matrix filling according to the aerial signal matrix and the ground signal matrix to obtain a recovered ground signal map.
Optionally, the method further includes:
the input module is used for recovering the signal of each area grid on the ground according to an aerial signal matrix based on a neural network model when the scene type of the signal area to be recovered is the first scene type, and inputting sample data into the neural network model before a recovered ground signal map is obtained; the sample data is an aerial signal matrix sample; the neural network model comprises two convolution layers, two pooling layers and a full-connection layer;
the output acquisition module is used for acquiring the recovered ground signal matrix output by the neural network model;
and the loss function updating module is used for updating the loss function of the neural network model according to the recovered ground signal matrix and the original ground signal matrix and continuing to train the neural network model.
Optionally, the second recovery module specifically includes:
the mapping unit is used for mapping the aerial signal matrix to the ground based on a channel fading model to obtain a ground signal mapping matrix; each nonzero element in the ground signal mapping matrix is a ground mapping signal of a corresponding area grid; the ground mapping signal is an average value of all aerial signals mapped to the area grid;
the fusion unit is used for fusing the ground signal mapping matrix and the ground signal matrix, determining the signal of each area grid and obtaining a ground signal updating matrix;
and the filling recovery unit is used for updating the matrix according to the ground signal and obtaining the recovered ground signal map by adopting a recovery method of accurate matrix filling.
Optionally, the fusion unit specifically includes:
the first judgment subunit is used for judging whether the ground signal corresponding to each area lattice is zero or not for each area lattice;
a first determining subunit, configured to determine, when the ground signal corresponding to the area lattice is not zero, the ground signal as a ground signal update value corresponding to the area lattice;
a second judging subunit, configured to, when the ground signal corresponding to the area lattice is zero, judge whether the ground mapping signal corresponding to the area lattice is zero;
a second determining subunit, configured to determine that a ground signal update value corresponding to the area lattice is zero when the ground mapping signal corresponding to the area lattice is zero;
a third determining subunit, configured to determine, when the ground mapping signal corresponding to the area lattice is not zero, the ground mapping signal as a ground signal update value corresponding to the area lattice;
and the ground signal updating matrix determining subunit is used for determining the ground signal updating matrix according to the ground signal updating values corresponding to all the area grids.
Optionally, the filling recovery unit specifically includes:
a decomposition subunit for converting the formula X into U-Sigma VTDecomposing the ground signal update matrix; wherein X is the ground signal update matrix, U is an MxM unitary matrix, Sigma is a singular value matrix, and VTIs an NxN unitary matrix;
a singular value matrix updating subunit, configured to reserve γ singular values, set zero to the singular values in the singular value matrix except for the reserved singular values, to obtain a singular value updating matrix Σγ；
A recovery subunit for updating the matrix according to the singular values using the formula Xγ＝UγΣγVγ TObtaining the recovered ground signal map Xγ(ii) a The recovered ground signal map XγUpdating an approximation matrix of a matrix X for the ground signal; wherein, UγA matrix consisting of the first k columns of elements of a unitary matrix of order U, Vγ TIs a unitary matrix V of orderTA matrix of the first k columns of elements.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
according to the method, the three-dimensional spatial correlation of the signals is considered, the signals have different correlations at different heights and different scenes, the ground signal map is quickly restored by adopting different restoration modes, and the restoration accuracy is improved. When the ground signal map is restored, the ground signal map can be well restored only by acquiring a small amount of ground signals or not acquiring the ground signals and assisting the ground signals with air signals, and the accuracy is high.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a method for recovering a ground signal map according to the present invention;
FIG. 2 is a schematic diagram of a neural network model according to the present invention;
FIG. 3 is a flow chart of recovery using a neural network model in the present invention;
FIG. 4 is a diagram illustrating the recovery effect of the neural network model according to the present invention;
FIG. 5 is a schematic illustration of the mapping of an aerial signal to a terrestrial signal in accordance with the present invention;
FIG. 6 is a schematic diagram of a terrestrial signal receiving aerial signal mapping in accordance with the present invention;
FIG. 7 is a schematic diagram illustrating a recovery process using a precise matrix filling method according to the present invention;
fig. 8 is a schematic structural diagram of the ground signal map recovery system according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a schematic flow chart of a ground signal map recovery method according to the present invention. As shown in fig. 1, the ground signal map recovery method of the present invention includes the steps of:
step 100: and acquiring an aerial signal matrix of a signal area to be recovered. The signal region to be restored includes M × N region lattices. The aerial signal matrix is a sparse matrix, the non-zero elements in the aerial signal matrix are aerial signals of grids in the corresponding area, and the zero elements in the aerial signal matrix indicate that the corresponding area grids do not acquire the aerial signals.
Step 200: and acquiring the scene type of the signal area to be recovered. The scene types comprise a first scene type and a second scene type, the signal fluctuation range in the first scene type is smaller than a first threshold value, the signal fluctuation range in the second scene type is larger than a second threshold value, and the first threshold value is far smaller than the second threshold value. And the values of the first threshold and the second threshold are adjusted according to the actual condition. The first scene type corresponds to scenes with small signal fluctuation, such as open ground and other scenes with small crowd density change; the second scene type corresponds to a scene with large signal fluctuation, such as a subway station, a shopping mall and other areas with large crowd density change. When the scene type of the signal area to be restored is the first scene type, executing step 300; when the scene type of the signal region to be restored is the second scene type, steps 400 and 500 are performed.
Step 300: when the scene type of the signal area to be recovered is the first scene type, recovering the signal of each area grid on the ground according to the aerial signal matrix based on the neural network model to obtain the recovered ground signalA map. Fig. 2 is a schematic structural diagram of a neural network model in the present invention, and fig. 3 is a flowchart of recovery using the neural network model in the present invention. Referring to fig. 2 and 3, the neural network model is composed of two convolutional layers, two pooling layers, and a fully connected layer. The input of the neural network model is a sparse air signal matrix, and the output is a complete ground signal matrix, namely a recovered ground signal map. Before the neural network model is adopted for recovery, a large number of sparse aerial signal matrix samples and corresponding original ground signal matrix training parameters are input into the neural network model, and a loss function of the neural network model is updated according to a recovered ground signal matrix and an original ground signal matrix output by each training, wherein the loss function is
Wherein the content of the first and second substances,
for the output recovered ground signal matrix,
is the original ground signal matrix. And then continuing to train the neural network model.
After training is finished, a sparse aerial signal matrix is input based on the trained neural network model, and then the recovered ground signal map can be obtained. The recovery effect is shown in fig. 4, and fig. 4 is an effect diagram of the recovery by using the neural network model in the present invention, in which, a part is a sparse aerial signal matrix schematic diagram, b part is an original ground signal matrix schematic diagram, and c part is a recovered ground signal map.
Step 400: and when the scene type of the signal area to be recovered is a second scene type, acquiring a ground signal matrix. And the non-zero elements in the ground signal matrix are ground signals of grids in the corresponding area.
Step 500: and recovering the signals of the grid of the area without the signal data on the ground by adopting a recovery method of accurate matrix filling according to the aerial signal matrix and the ground signal matrix to obtain a recovered ground signal map. Fig. 7 is a schematic flow chart of recovery by the recovery method using accurate matrix filling according to the present invention, and as shown in fig. 7, the specific process is as follows:
(1) and mapping the aerial signal matrix to the ground based on a channel fading model to obtain a ground signal mapping matrix. Inspired by channel fading model
P represents the received power, PtDenotes transmission power, GtRepresenting transmission gain, GrThe receiving gain is represented, lambda represents the wavelength, d is the distance from the air area grid to the ground area grid, the air area grid and the ground area grid are in one-to-one correspondence, and L is the system loss. By approximating the signal point in the air to the transmitting base station, the ground signal point centered on the signal point in the air also approximately conforms to the model, i.e., the inverse square distance relationship
Wherein SijA signal representing the mapping of the air signal to the ground area grid, c being a constant, S'ijIs a signal of an air area grid, namely an air signal. For one air region lattice l (i ', j') and ground region lattice l (i, j), the k constraint is defined as follows:
that is, for an air signal point, assuming that it can be mapped to a grid on the ground centered on it with a distance not exceeding k, one air signal point can be mapped to 2(k +1) in an area divided into several area grids2Fig. 5 is a schematic diagram illustrating the mapping of the aerial signal to the ground signal according to the present invention. Similarly, for a signal point on the ground, it will also receive a signal from the air 2(k +1)2Mapping of individual signal points, as shown in FIG. 6, FIG. 6 is a schematic diagram of the mapping of terrestrial signal reception aerial signals of the present invention。
Further, 2(k +1) received in each ground area cell can be obtained2A mapping value of the aerial signal, 2(k +1)2And taking weighted average of the mapping values as the ground mapping signal of the region grid to obtain a ground signal mapping matrix. Each non-zero element in the ground signal mapping matrix is a ground mapping signal of a corresponding area grid, and a zero element indicates that no air signal is mapped to the ground in the area grid.
(2) And fusing the ground signal mapping matrix and the ground signal matrix, determining the signal of each area grid, and obtaining a ground signal updating matrix.
If the signal mapped to the ground by the aerial signal cannot well recover the complete signal, a small amount of signals need to be collected on the ground for assistance, so that the final matrix to be recovered is formed by the mapped signal and the signal collected on the ground. For a signal of an area grid on the ground, if the signal of the area grid is collected by ground crowdsourcing participants, the signal is used as the signal strength of the area grid, otherwise, a ground mapping signal mapped by an air signal is used as the signal of the area grid. This embodiment uses a 0-1 matrix IM×NTo indicate whether the current lattice point is selected as a sampling point, when l (i, j) is selected as a sampling point, the value of the corresponding area lattice is 1; when l (i, j) is not selected as a sampling point, the value of the corresponding area lattice is 0.
The signal strength x of the grid for a certain area on the groundijAnd satisfies the following conditions:
wherein, gSijThe ground signal intensity in the area grid l (i, j) collected on the ground; i isi,jRepresentation matrix IM×NRow I and column j ofi,j1 denotes that the ground grid l (I, j) is selected as a sampling point, Ii,j0 means that the ground grid l (i, j) is not selected as a sampling point;
the mth cell l (i) representing the aerial regionm'jm') distance from the ground area grid l (i, j);
the nth cell l (i) representing the sky arean'jn') distance from the ground area grid l (i, j);
the m-th cell l (i) of the aerial regionm'jm') is selected from the group consisting of,
the mth cell l (i) representing the aerial regionm'jm') is selected as the sampling point,
the mth cell l (i) representing the aerial regionm'jm') not selected as a sampling point;
for the nth cell l (i) of the sky arean'jn') is selected from the group consisting of,
the nth cell l (i) representing the sky arean'jn') is selected as the sampling point,
the nth cell l (i) representing the sky arean'jn') not selected as a sampling point;
denotes the m cells l (i'm,j′m) The signal strength of (c).
For each area grid on the ground, there are four cases where the signal values originate:
mapping derived only from the signals in the air: the signal value of the ground area grid is calculated according to the formula, namely, the ground mapping signal is determined as the ground signal corresponding to the area grid.
Derived only from the ground participant perception: the signal value of the ground grid point is a value collected by a ground participant, namely the ground signal is determined as the ground signal corresponding to the grid of the area.
The mapping of the air signals and the acquisition of the ground participants are available: considering that the signal precision acquired by the ground is higher than the signal value obtained by mapping, the signal value of the ground area grid is based on the value acquired by the ground participant, i.e. the ground signal is determined as the ground signal corresponding to the area grid.
There is neither mapping of the air signals nor ground participant acquisition: and setting the grid signal value to zero, namely setting the ground signal corresponding to the grid of the area to zero.
(3) And updating the matrix according to the ground signals, and obtaining the recovered ground signal map by adopting a recovery method of accurate matrix filling. The present invention uses an approximate rank to participate in the computation. The specific process is as follows:
decomposing the ground signal update matrix into X ═ U Σ V by singular value decompositionT(ii) a Wherein X is the ground signal update matrix, U is an MxM unitary matrix, Sigma is a singular value matrix, and VTIs an N × N unitary matrix.
Then, only gamma singular values are reserved, other singular values except the reserved singular values in the singular value matrix are set to be zero, and a singular value updating matrix sigma is obtainedγ；
Finally, the matrix is updated according to the singular values, using formula Xγ＝UγΣγVγ TCalculating to obtain an approximate matrix X of the ground signal updating matrix XγNamely, the recovered ground signal map is obtained. Wherein, UγA matrix consisting of the first k columns of elements of a unitary matrix of order U, Vγ TIs a unitary matrix V of orderTA matrix of the first k columns of elements.
The scheme of the embodiment is based on a propagation model, maps signal points in the air to a plurality of signal points on the ground, and finally recovers the signal map by using a matrix filling mode.
Corresponding to the method for recovering the ground signal map shown in fig. 1, the present invention further provides a system for recovering the ground signal map, and fig. 8 is a schematic structural diagram of the system for recovering the ground signal map according to the present invention. As shown in fig. 8, the ground signal map recovery system of the present invention includes the following structures:
an aerial signal matrix obtaining module 801, configured to obtain an aerial signal matrix; the aerial signal matrix is a sparse matrix, and the non-zero elements in the aerial signal matrix are aerial signals of corresponding region lattices.
A scene type obtaining module 802, configured to obtain a scene type of a signal region to be recovered; the scene types comprise a first scene type and a second scene type; the signal fluctuation range in the first scene type is smaller than a first threshold, the signal fluctuation range in the second scene type is larger than a second threshold, and the first threshold is smaller than the second threshold.
A first recovery module 803, configured to recover, based on the neural network model, a signal of each grid of the ground area according to the aerial signal matrix when the scene type of the signal area to be recovered is the first scene type, so as to obtain a recovered ground signal map.
A ground signal matrix obtaining module 804, configured to obtain a ground signal matrix when the scene type of the signal area to be recovered is the second scene type; and the non-zero elements in the ground signal matrix are ground signals of grids in the corresponding area.
A second recovery module 805, configured to recover, according to the aerial signal matrix and the ground signal matrix, a signal of a grid in an area without signal data on the ground by using a recovery method of accurate matrix filling, so as to obtain a recovered ground signal map.
As another embodiment, the ground signal map recovery system of the present invention further includes:
the input module is used for recovering the signal of each area grid on the ground according to an aerial signal matrix based on a neural network model when the scene type of the signal area to be recovered is the first scene type, and inputting sample data into the neural network model before a recovered ground signal map is obtained; the sample data is an aerial signal matrix sample; the neural network model comprises two convolution layers, two pooling layers and a full-connection layer;
the output acquisition module is used for acquiring the recovered ground signal matrix output by the neural network model;
and the loss function updating module is used for updating the loss function of the neural network model according to the recovered ground signal matrix and the original ground signal matrix and continuing to train the neural network model.
As another embodiment, in the ground signal map recovery system according to the present invention, the second recovery module 805 specifically includes:
the mapping unit is used for mapping the aerial signal matrix to the ground based on a channel fading model to obtain a ground signal mapping matrix; each nonzero element in the ground signal mapping matrix is a ground mapping signal of a corresponding area grid; the ground mapping signal is an average value of all aerial signals mapped to the area grid;
the fusion unit is used for fusing the ground signal mapping matrix and the ground signal matrix, determining the signal of each area grid and obtaining a ground signal updating matrix;
and the filling recovery unit is used for updating the matrix according to the ground signal and obtaining the recovered ground signal map by adopting a recovery method of accurate matrix filling.
As another embodiment, in the ground signal map recovery system of the present invention, the fusion unit specifically includes:
the first judgment subunit is used for judging whether the ground signal corresponding to each area lattice is zero or not for each area lattice;
a first determining subunit, configured to determine, when the ground signal corresponding to the area lattice is not zero, the ground signal as a ground signal update value corresponding to the area lattice;
a second judging subunit, configured to, when the ground signal corresponding to the area lattice is zero, judge whether the ground mapping signal corresponding to the area lattice is zero;
a second determining subunit, configured to determine that a ground signal update value corresponding to the area lattice is zero when the ground mapping signal corresponding to the area lattice is zero;
a third determining subunit, configured to determine, when the ground mapping signal corresponding to the area lattice is not zero, the ground mapping signal as a ground signal update value corresponding to the area lattice;
and the ground signal updating matrix determining subunit is used for determining the ground signal updating matrix according to the ground signal updating values corresponding to all the area grids.
As another embodiment, in the ground signal map recovery system according to the present invention, the filling recovery unit specifically includes:
a decomposition subunit for converting the formula X into U-Sigma VTDecomposing the ground signal update matrix; wherein X is the ground signal update matrix, U is an MxM unitary matrix, Sigma is a singular value matrix, and VTIs an NxN unitary matrix;
the singular value matrix updating subunit is used for reserving gamma singular values and setting the singular values in the singular value matrix except the reserved singular values to zero to obtain a singular value updating matrix;
a recovery subunit for updating the matrix according to the singular values using the formula Xγ＝UγΣγVγ TObtaining the recovered ground signal map Xγ(ii) a The recovered ground signal map XγUpdating an approximation matrix of a matrix X for the ground signal; wherein, UγA matrix consisting of the first k columns of elements of a unitary matrix of order U, Vγ TIs a unitary matrix V of orderTFirst k columnsA matrix of elements.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (2)
1. A method for recovering a ground signal map, comprising:
acquiring an aerial signal matrix of a signal area to be recovered; the signal area to be recovered comprises M multiplied by N area lattices; the aerial signal matrix is a sparse matrix, and nonzero elements in the aerial signal matrix are aerial signals of corresponding region lattices;
acquiring the scene type of the signal area to be recovered; the scene types comprise a first scene type and a second scene type; the signal fluctuation range in the first scene type is smaller than a first threshold value, the signal fluctuation range in the second scene type is larger than a second threshold value, and the first threshold value is smaller than the second threshold value;
when the scene type of the signal area to be recovered is the first scene type, recovering the signal of each grid of the ground area according to an aerial signal matrix based on a neural network model to obtain a recovered ground signal map; before the signal of each grid of each area on the ground is recovered, the method further comprises the following steps: inputting sample data into the neural network model; the sample data is an aerial signal matrix sample; the neural network model comprises two convolution layers, two pooling layers and a full-connection layer; acquiring a recovered ground signal matrix output by the neural network model; updating the loss function of the neural network model according to the recovered ground signal matrix and the original ground signal matrix, and continuing to train the neural network model;
when the scene type of the signal area to be recovered is the second scene type, acquiring a ground signal matrix; non-zero elements in the ground signal matrix are ground signals of grids in the corresponding area;
according to the aerial signal matrix and the ground signal matrix, recovering signals of the grid of the region without signal data on the ground by adopting a recovery method of accurate matrix filling to obtain a recovered ground signal map; the method specifically comprises the following steps:
mapping the aerial signal matrix to the ground based on a channel fading model to obtain a ground signal mapping matrix; each nonzero element in the ground signal mapping matrix is a ground mapping signal of a corresponding area grid; the ground mapping signal is an average value of all aerial signals mapped to the area grid;
fusing the ground signal mapping matrix and the ground signal matrix, and determining a signal of each area grid to obtain a ground signal updating matrix; the specific process is as follows: for each area lattice, judging whether the ground signal corresponding to the area lattice is zero or not; when the ground signal corresponding to the area grid is not zero, determining the ground signal as a ground signal update value corresponding to the area grid; when the ground signal corresponding to the area grid is zero, judging whether the ground mapping signal corresponding to the area grid is zero or not; when the ground mapping signal corresponding to the area lattice is zero, determining that the ground signal updating value corresponding to the area lattice is zero; when the ground mapping signal corresponding to the area grid is not zero, determining the ground mapping signal as a ground signal updating value corresponding to the area grid; determining a ground signal updating matrix according to ground signal updating values corresponding to all the area grids;
according to the ground letterUpdating the matrix by the number, and obtaining a recovered ground signal map by adopting a recovery method of accurate matrix filling; the specific process is as follows: according to the formula X ═ U ∑ VTDecomposing the ground signal update matrix; wherein X is the ground signal update matrix, U is an MxM unitary matrix, Sigma is a singular value matrix, and VTIs an NxN unitary matrix; reserving gamma singular values, setting the singular values except the reserved singular values in the singular value matrix to zero to obtain a singular value updating matrix sigmaγ(ii) a Updating the matrix according to the singular values, using formula Xγ＝UγΣγVγ TObtaining the recovered ground signal map Xγ(ii) a The recovered ground signal map XγUpdating an approximation matrix of a matrix X for the ground signal; wherein, UγA matrix consisting of the first k columns of elements of a unitary matrix of order U, Vγ TIs a unitary matrix V of orderTA matrix of the first k columns of elements.
2. A ground signal map recovery system, comprising:
the aerial signal matrix acquisition module is used for acquiring an aerial signal matrix of a signal area to be recovered; the signal area to be recovered comprises M multiplied by N area lattices; the aerial signal matrix is a sparse matrix, and nonzero elements in the aerial signal matrix are aerial signals of corresponding region lattices;
a scene type obtaining module, configured to obtain a scene type of the signal region to be restored; the scene types comprise a first scene type and a second scene type; the signal fluctuation range in the first scene type is smaller than a first threshold value, the signal fluctuation range in the second scene type is larger than a second threshold value, and the first threshold value is smaller than the second threshold value;
the first recovery module is used for recovering the signal of each area grid on the ground according to an aerial signal matrix based on a neural network model when the scene type of the signal area to be recovered is the first scene type to obtain a recovered ground signal map;
a ground signal matrix obtaining module, configured to obtain a ground signal matrix when the scene type of the signal area to be recovered is the second scene type; non-zero elements in the ground signal matrix are ground signals of grids in the corresponding area;
the second recovery module is used for recovering the signals of the grid in the area without the signal data on the ground by adopting a recovery method of accurate matrix filling according to the aerial signal matrix and the ground signal matrix to obtain a recovered ground signal map; the second recovery module specifically includes:
the mapping unit is used for mapping the aerial signal matrix to the ground based on a channel fading model to obtain a ground signal mapping matrix; each nonzero element in the ground signal mapping matrix is a ground mapping signal of a corresponding area grid; the ground mapping signal is an average value of all aerial signals mapped to the area grid;
the fusion unit is used for fusing the ground signal mapping matrix and the ground signal matrix, determining the signal of each area grid and obtaining a ground signal updating matrix; the fusion unit specifically includes: the first judgment subunit is used for judging whether the ground signal corresponding to each area lattice is zero or not for each area lattice; a first determining subunit, configured to determine, when the ground signal corresponding to the area lattice is not zero, the ground signal as a ground signal update value corresponding to the area lattice; a second judging subunit, configured to, when the ground signal corresponding to the area lattice is zero, judge whether the ground mapping signal corresponding to the area lattice is zero; a second determining subunit, configured to determine that a ground signal update value corresponding to the area lattice is zero when the ground mapping signal corresponding to the area lattice is zero; a third determining subunit, configured to determine, when the ground mapping signal corresponding to the area lattice is not zero, the ground mapping signal as a ground signal update value corresponding to the area lattice; the ground signal updating matrix determining subunit is used for determining the ground signal updating matrix according to the ground signal updating values corresponding to all the area grids;
the recovery unit is filled in with the liquid,the recovery method is used for updating the matrix according to the ground signal and adopting accurate matrix filling to obtain a recovered ground signal map; the filling recovery unit specifically includes: a decomposition subunit for converting the formula X into U-Sigma VTDecomposing the ground signal update matrix; wherein X is the ground signal update matrix, U is an MxM unitary matrix, Sigma is a singular value matrix, and VTIs an NxN unitary matrix; a singular value matrix updating subunit, configured to reserve γ singular values, set zero to the singular values in the singular value matrix except for the reserved singular values, to obtain a singular value updating matrix Σγ(ii) a A recovery subunit for updating the matrix according to the singular values using the formula Xγ＝UγΣγVγ TObtaining the recovered ground signal map Xγ(ii) a The recovered ground signal map XγUpdating an approximation matrix of a matrix X for the ground signal; wherein, UγA matrix consisting of the first k columns of elements of a unitary matrix of order U, Vγ TIs a unitary matrix V of orderTA matrix composed of the first k columns of elements;
the ground signal map recovery system further comprises:
the input module is used for recovering the signal of each area grid on the ground according to an aerial signal matrix based on a neural network model when the scene type of the signal area to be recovered is the first scene type, and inputting sample data into the neural network model before a recovered ground signal map is obtained; the sample data is an aerial signal matrix sample; the neural network model comprises two convolution layers, two pooling layers and a full-connection layer;
the output acquisition module is used for acquiring the recovered ground signal matrix output by the neural network model;
and the loss function updating module is used for updating the loss function of the neural network model according to the recovered ground signal matrix and the original ground signal matrix and continuing to train the neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911326197.7A CN111065123B (en) | 2019-12-20 | 2019-12-20 | Ground signal map recovery method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911326197.7A CN111065123B (en) | 2019-12-20 | 2019-12-20 | Ground signal map recovery method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111065123A CN111065123A (en) | 2020-04-24 |
| CN111065123B true CN111065123B (en) | 2021-09-14 |
Family
ID=70301296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911326197.7A Active CN111065123B (en) | 2019-12-20 | 2019-12-20 | Ground signal map recovery method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111065123B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113030569A (en) * | 2021-02-26 | 2021-06-25 | 华南理工大学 | Line spectrum estimation method |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10032538B2 (en) * | 2013-11-13 | 2018-07-24 | The United States Of America As Represented By The Secretary Of The Army | Deformable elastomeric conductors and differential electronic signal transmission |
| CN105512646B (en) * | 2016-01-19 | 2019-03-01 | 腾讯科技（深圳）有限公司 | A kind of data processing method, device and terminal |
| CN109068387B (en) * | 2018-09-19 | 2020-05-12 | 中国科学技术大学 | WSN positioning architecture integrating communication and calculation and implementation method |
| CN109861738B (en) * | 2018-12-26 | 2020-04-07 | 西安电子科技大学 | Method for realizing asynchronous transmission in parallel QR decomposition relay satellite forwarding system |
-
2019
- 2019-12-20 CN CN201911326197.7A patent/CN111065123B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111065123A (en) | 2020-04-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101216304B (en) | Systems and methods for object dimension estimation | |
| CN112632311B (en) | Cloud layer change trend prediction method based on deep learning | |
| CN106687773A (en) | System and methods for sensor node localization and sensor network organization based on contextual event detection | |
| CN105160678A (en) | Convolutional-neural-network-based reference-free three-dimensional image quality evaluation method | |
| CN108828643B (en) | Indoor and outdoor seamless positioning system and method based on grey prediction model | |
| CN111539453B (en) | Global ionized layer electron total content prediction method based on deep cycle neural network | |
| CN113837361A (en) | Air pollutant concentration prediction method and system | |
| CN110232471A (en) | A kind of precipitation sensing net node layout optimization method and device | |
| CN113095437B (en) | Fire point detection method for Himapari-8 remote sensing data | |
| CN111065123B (en) | Ground signal map recovery method and system | |
| CN115393712B (en) | SAR image road extraction method and system based on dynamic hybrid pooling strategy | |
| JP2017228091A (en) | Walking state learning method, walking state estimation method, road surface condition comprehension method, device and program | |
| CN106845343A (en) | A kind of remote sensing image offshore platform automatic testing method | |
| CN114912707A (en) | Air quality prediction system and method based on multi-mode fusion | |
| CN112906046A (en) | Model training method and device by using single-bit compression perception technology | |
| CN117214904A (en) | Intelligent fish identification monitoring method and system based on multi-sensor data | |
| CN116106988A (en) | Weather prediction method and device, electronic equipment and storage medium | |
| CN113887330A (en) | Target detection system based on remote sensing image | |
| CN117710508A (en) | Near-surface temperature inversion method and device for generating countermeasure network based on improved condition | |
| CN113553949A (en) | Tailing pond semantic segmentation method based on photogrammetric data | |
| CN113989612A (en) | Remote sensing image target detection method based on attention and generation countermeasure network | |
| CN112380945A (en) | Method and system for identifying potential environmental hazards of power transmission channel | |
| Xu et al. | UAV-ODS: A real-time outfall detection system based on UAV remote sensing and edge computing | |
| CN115561834A (en) | Meteorological short-term and temporary forecasting all-in-one machine based on artificial intelligence | |
| CN116310828A (en) | High-resolution remote sensing image change detection method and device combining transducer and CNN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |