A method and device for detecting adjacent hidden human targets based on time-space fusion

Abstract

The present invention discloses a method and device for detecting adjacent hidden human targets based on time-space fusion, and belongs to the field of low-frequency radar imaging technology. The method comprises processing the target echo collected by the radar using non-causal moving target indication, removing the stationary target echo, and obtaining an effective multi-channel echo signal; performing coherent imaging on the effective multi-channel echo signal to obtain multiple frames of images; using the obtained multiple frames of images to build a preliminary time-space feature model; optimizing the preliminary time-space feature model to obtain a high signal-to-noise ratio time-space feature. The present invention can fuse the time and space resolution of the radar, greatly improve the accuracy of adjacent target detection, and realize the simultaneous acquisition of the spatial position and respiratory frequency of the human target.

Description

Technical Field

[0001] The present invention belongs to the technical field of low-frequency radar imaging, and in particular relates to a method and a device for detecting adjacent hidden human targets based on time-space fusion. Background Art

[0002] Low-frequency electromagnetic waves can penetrate obstacles to detect human targets behind them, and are widely used in many fields such as earthquake relief. The existing technology uses a radar array to transmit and receive low-frequency electromagnetic waves. First, based on the micro-motion characteristics of human breathing, moving target indication is used to separate it from the static target echo. Among them, the moving target indication includes background subtraction and adjacent subtraction. Further, radar imaging technology (mainly based on the back projection algorithm) is used to obtain the reflection coefficient image of the detection area, that is, the spatial image, and the bright spots therein are identified as human targets to obtain their spatial positions.

[0003] Existing related research is based on synthetic aperture (SAR) arrays and multiple-transmit multiple-receive (MIMO) real aperture arrays. Their ability to distinguish multiple human targets depends on the spatial resolution of the radar (including range and azimuth resolution). Only when the distance between targets is greater than the spatial resolution of the radar can they be distinguished, otherwise they will be detected as the same target. In practice, the spatial resolution of the radar is limited, and adjacent hidden human targets cannot be identified, which seriously reduces the accuracy of target detection. Therefore, how to break through the limitation of radar resolution and improve the detection capability of adjacent targets has become the focus of research. Summary of the invention

[0004] To solve the above technical problems, the present invention provides a method and device for detecting adjacent hidden human targets based on time-space fusion, which adopts non-causal moving target indication. During imaging, the frequency dimension of a single-channel echo signal is first converted into an intuitive distance, and the spatial and temporal dimension information of multiple frames of images is integrated. The process of building a time-space feature model is derived, which has a better effect of removing stationary target echoes, avoids the occurrence of large-scale matrix calculations in the optimization process, and obtains time-space features with high signal-to-noise ratio through optimization.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] A method for detecting adjacent hidden human targets based on spatiotemporal fusion, comprising:

[0007] Step 1: Use non-causal moving target indication to process the target echo collected by the radar, remove the static target echo, and obtain an effective multi-channel echo signal;

[0008] Step 2: performing coherent imaging on the effective multi-channel echo signals to obtain multiple frames of images;

[0009] Step 3: Use the acquired multi-frame images to build a preliminary spatiotemporal feature model;

[0010] Step 4: Construct forward space-time sensing factors and backward space-time sensing factors based on the preliminary space-time feature model, construct and solve the constrained optimization objective function, and obtain space-time features with high signal-to-noise ratio.

[0011] In a second aspect, the present invention provides an adjacent hidden human target detection device based on time-space fusion, comprising:

[0012] A data preprocessing unit is used to process the target echo collected by the radar using non-causal moving target indication, remove the stationary target echo, and obtain an effective multi-channel echo signal;

[0013] An image acquisition unit, used for performing coherent imaging on effective multi-channel echo signals to acquire multiple frames of images;

[0014] A model building unit, used to build a preliminary spatiotemporal feature model using the acquired multiple frames of images;

[0015] The spatiotemporal feature acquisition unit is used to construct forward spatiotemporal sensing factors and backward spatiotemporal sensing factors based on the preliminary spatiotemporal feature model, construct and solve the constrained optimization objective function, and obtain spatiotemporal features with a high signal-to-noise ratio.

[0016] In a third aspect, the present invention provides an electronic device comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned adjacent hidden human target detection method based on spatiotemporal fusion.

[0017] In a fourth aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned adjacent hidden human target detection method based on spatiotemporal fusion.

[0018] The beneficial effects of the present invention are:

[0019] By integrating the time dimension and the space dimension, adjacent targets can be distinguished not only from the spatial perspective but also from the temporal perspective, which greatly improves the accuracy of adjacent target detection.

[0020] The final result includes the direction, distance and frequency at the same time, which enables the simultaneous acquisition of the spatial position and breathing frequency of the human target.

[0021] The proposed optimized part can effectively suppress azimuth sidelobes, background noise, target artifacts and frequency-doubled signals.

[0022] It is not limited by radar system and array form, and is applicable to various situations such as pulse system, frequency modulated continuous wave system and stepped frequency radar system, and can be effectively applied to various array forms such as linear array, uniform array and planar array. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a method for detecting adjacent hidden human targets based on time-space fusion according to the present invention;

[0024] Figure 2 A simulation model diagram of an embodiment of the present invention;

[0025] Figure 3 The antenna array position diagram used in the simulation model of the embodiment of the present invention;

[0026] Figure 4 A single frame echo signal curve diagram of the first channel simulated in an embodiment of the present invention;

[0027] Figure 5 A multi-frame data diagram of the first channel simulated by an embodiment of the present invention;

[0028] Figure 6 A single-frame imaging result diagram of an embodiment of the present invention;

[0029] Figure 7a It is a three-dimensional display diagram of the spatiotemporal feature model;

[0030] Figure 7b It is a two-dimensional projection display of the azimuth-distance of the spatiotemporal feature model;

[0031] Figure 7c It is a two-dimensional projection display of the azimuth-frequency of the spatiotemporal characteristic model;

[0032] Figure 7d It is a two-dimensional projection display of distance-frequency for the spatiotemporal feature model;

[0033] Figure 8a This is a three-dimensional display diagram of the optimized spatiotemporal feature model;

[0034] Figure 8b To display the azimuth-distance two-dimensional projection of the optimized spatiotemporal feature model;

[0035] Figure 8c It is the azimuth-frequency two-dimensional projection display of the optimized spatiotemporal feature model;

[0036] Figure 8d This is the distance-frequency two-dimensional projection display of the optimized spatiotemporal feature model. DETAILED DESCRIPTION

[0037] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0038] like Figure 1 As shown, it is a flow chart of a method for detecting adjacent hidden human targets based on time-space fusion of the present invention, which includes the following steps:

[0039] Step 1: Use non-causal moving target indication to process the target echo collected by the radar, remove the static target echo, and obtain effective multi-channel echo signals;

[0040] Step 2: Perform coherent imaging on the effective multi-channel echo signals to obtain multiple frames of images;

[0041] Step 3: Use the acquired multi-frame images to build a preliminary spatiotemporal feature model;

[0042] Step 4: Optimize the preliminary spatiotemporal feature model to improve the signal-to-noise ratio and obtain spatiotemporal features with high signal-to-noise ratio, such as the spatial coordinates of human targets and breathing frequency, including constructing forward spatiotemporal sensing factors and backward spatiotemporal sensing factors based on the preliminary spatiotemporal feature model, constructing a constrained optimization objective function and solving it to obtain spatiotemporal features with high signal-to-noise ratio.

[0043] Specifically, step 1: using non-causal moving target indication to process the target echo collected by the radar, and removing the stationary target echo is implemented as follows;

[0044] The radar array used is multiple transmit multiple receive (MIMO), where the number of transmitting antennas is , the number of receiving antennas is Therefore, the total number of channels is The transmitting antenna transmits a frequency modulated continuous wave signal, which is reflected by the target and received by the receiving antenna. This reflected echo contains not only the reflected echo of the human target, but also the reflected echo of other stationary targets. Their energy is much higher than the reflection of the human target, and they are removed by non-causal moving target indication. Different from the conventional moving target indication method, the non-causal moving target indication in the present invention combines the past and the future, which can more effectively remove the stationary target echo. The original reflected echo signal of each frame is recorded as , three dimensions , and Represent the transmit antenna index, receive antenna index and frequency index respectively, and their maximum values ​​are , and Arrange multiple frames of original echo signals in chronological order, record the first The original echo signal of the frame is , the removal of stationary target echo can be written as:

[0045] (1)

[0046] in, Representative Frame original echo signal, After removing the stationary target echo, Frame signal, Indicates the number of background frames. is the total number of frames after the stationary target echo is removed. At this time, only the reflected echo of the human target remains, so It can also be expressed as:

[0047] (2)

[0048] in, Represents the position of the human target, which is composed of the initial position of the human target and chest wall rise and fall Superimposed, Represents the slow time dimension. is the reflection coefficient of the human target, and are the starting frequency and frequency modulation slope of the transmitted frequency modulated continuous wave (FMCW) signal, is the sampling time interval, is the time delay of the received signal relative to the transmitted signal. Recorded as .

[0049] Step 2: Perform coherent imaging on the signal processed in step 1 to obtain multiple frames of images including:

[0050] In order to avoid the large amount of calculation caused by the conventional back-projection (BP) imaging algorithm, its frequency dimension superposition operation is transformed. Frame signal For example, the frequency dimension of a single-channel echo signal is converted into an intuitive distance display:

[0051] (3) in, Representatives along Dimensions are subjected to inverse fast Fourier transform (IFFT), and K are the distance index and its maximum value respectively.

[0052] The imaging area is evenly divided into rectangular grids, and the signal is calculated from the transmitting antenna to the grid points. And then the time delay back to the receiving antenna:

[0053] (4)

[0054] in, and are the positions of the transmitting antenna and the receiving antenna, respectively. Represents the propagation speed of electromagnetic waves.

[0055] The pixel values ​​of each grid point are combined into the Frame imaging results. For example, its pixel value can be obtained by coherently superimposing the echoes of each channel:

[0056] (5)

[0057] in, represent exist The value of Represents the center frequency, and its expression is:

[0058] (6)

[0059] Repeat the above steps for each frame of echo data to obtain multiple frames of images.

[0060] Step 3: Use the multi-frame images obtained in step 2 to preliminarily build a spatiotemporal feature model;

[0061] When the grid point Exactly at the initial position of the human target hour, and The difference can be simplified:

[0062] (7)

[0063] in, represents the 2-norm operation, that is To obtain The 2-norm of .

[0064] Therefore, the grid points The pixel values ​​can be expanded along the time dimension:

[0065] (8)

[0066] The rise and fall of the chest wall Modeled as a sinusoidal wave:

[0067] (9)

[0068] in, Represents the amplitude of chest wall rise and fall, Represents the frequency of chest wall rise and fall.

[0069] Further:

[0070] (10)

[0071] in, represent Bessel series of the first kind.

[0072] right By performing spectrum analysis, we can get:

[0073] (11)

[0074] in, Representatives along Dimensions are subjected to a fast Fourier transform (FFT), represents the frequency index in the slow time direction, represents the Dirac function.

[0075] At this time, it can be observed that the energy in the slow time direction is distributed along integer multiples of the respiratory frequency. hour, Only the breathing features are included. Expanded to a comprehensive representation of all pixels:

[0076] (12)

[0077] (13)

[0078] in, Representative Frame sub-image, Represents the matrix transpose operation. middle, Dimensions provide spatial information. It provides slow time information. For human targets, its spatial position and breathing frequency can be obtained at the same time, thus completing the preliminary construction of the spatiotemporal feature model.

[0079] Step 4: Optimize the spatiotemporal feature model of step 3 to obtain spatiotemporal features with high signal-to-noise ratio.

[0080] In practice, the reflection signal of the human target is very weak, which leads to a low signal-to-noise ratio. Considering the influence of factors such as the involuntary micro-motion of the human trunk, it is difficult to extract the spatial position and breathing frequency of the human target from the initial spatiotemporal feature model obtained in step 3. Therefore, the initial spatiotemporal feature model is further optimized to improve the signal-to-noise ratio.

[0081] In order to avoid the occurrence of large-scale matrix calculations, a forward and backward spatiotemporal sensing factors :

[0082] (14)

[0083] The parameter expressions involved are:

[0084] (15)

[0085] and, and Respectively represent expanding the matrix in the brackets to four dimensions and five dimensions to satisfy the Hadamard product requirements. , The intermediate parameters and The conjugate of is the Fourier transform matrix, , is its transpose and conjugate transpose form.

[0086] Since the distribution of the spatial position and breathing frequency of the human body target is sparse, using this sparse characteristic, the optimization objective can be written as:

[0087] (16)

[0088] in, represents the 1-norm operation, represents the 2-norm operation, is a regularization parameter used to balance the accuracy and sparsity of the optimization results. , the above unconstrained optimization objective can be converted into a constrained optimization objective:

[0089] (17)

[0090] From this we can get the matrix form of the augmented Lagrangian function:

[0091] (18)

[0092] in, and is the Lagrange multiplier and its conjugate transpose, represents the trace calculation of the matrix, is the penalty parameter.

[0093] The optimization problem of formula (18) can be converted into three sub-optimization problems. After derivation, their iterative solutions are:

[0094] (19)

[0095] in, The index of the iteration number. is the soft threshold function, and its calculation formula is:

[0096] (20)

[0097] The optimization result can be obtained by using the above iterations.

[0098] Example

[0099] In the embodiment, specific simulation parameters are set and the open source gprMax software is used to simulate and analyze the feasibility of the present invention. The simulation model is shown in the attached figure. Figure 2 As shown, two cylinders with sinusoidally varying radii are used to simulate human breathing.

[0100] The antenna array settings of the radar in the simulation model are shown in the attached figure. Figure 3 As shown, the antenna array is evenly arranged parallel to Axis placement.

[0101] The simulation parameter settings are shown in Table 1:

[0102] Table 1

[0103]

[0104] Taking the center of the array as the coordinate origin, the coordinates and micro-motion frequencies of the two targets are:

[0105] Target 1: (-0.15, 0, 4.5), 0.22Hz

[0106] Target 2: (0.15, 0, 4.5), 0.18Hz

[0107] Select the single frame echo signal of the first channel, such as Figure 4 As shown in Figure 2, the direct wave, obstacle reflection and secondary reflection are much stronger than the target reflection.

[0108] After the stationary target echo is removed, the multi-frame data of the first channel are plotted in a graph, such as Figure 5 As shown in the figure. At this time, only the micro-moving cylindrical target echo remains, and the static target echo has been removed. However, only a single micro-movement curve can be clearly seen in the figure, which means that the two micro-moving targets cannot be distinguished. In addition, the single-channel echo signal only shows the direct distance from the target to the radar, and the spatial position of the target cannot be extracted.

[0109] Using the multi-channel echo signal after removing the stationary target echo, imaging is performed according to step 2, and the obtained multi-frame images are basically the same. Figure 6 One frame of the image is shown.

[0110] It can be seen that after imaging, the signal-to-noise ratio has been improved to a certain extent compared with single-channel, and the spatial position of the target has been obtained. But most importantly, the imaging result also only shows a single target, which means that the two targets are still indistinguishable.

[0111] Furthermore, using the content described in step 3, a preliminary spatiotemporal feature model is constructed. Its three-dimensional results and its projections along three dimensions are shown in Figures 7a-7d As shown, Figure 7a .3D display at -18dB dynamic range; Figure 7b . Azimuth-distance two-dimensional projection display; Figure 7c .Azimuth-frequency two-dimensional projection display; Figure 7d .Distance-frequency two-dimensional projection display.

[0112] As can be seen from the initial spatiotemporal feature model in Figure 7, the two targets can be distinguished. However, there is a lot of interference at this time, mainly including: azimuth side lobes, artifacts behind the target, frequency-doubled signals, and background noise, which makes it impossible to correctly extract the target information.

[0113] Using the optimization described in step 4, the optimized spatiotemporal features are obtained as follows: Figures 8a-8d As shown, Figure 8a .3D display at -18dB dynamic range; Figure 8b . Azimuth-distance two-dimensional projection display; Figure 8c .Azimuth-frequency two-dimensional projection display; Figure 8d .Distance-frequency two-dimensional projection display.

[0114] Depend on Figures 8a-8d It can be seen that, on the one hand, the two targets have achieved a good resolution effect; on the other hand, the interference has been completely suppressed under the same dynamic range as the preliminary spatiotemporal feature model. The spatial coordinates and micro-motion frequencies of the two targets obtained are:

[0115] Target 1: (-0.23, 0, 4.63), 0.22Hz

[0116] Target 2: (0.14, 0, 4.63), 0.18Hz

[0117] This is consistent with the target information set during simulation, thus proving the feasibility of the present invention.

[0118] In summary, different from the conventional moving target indication method, the present invention adopts non-causal moving target indication, which has a better effect of removing stationary target echoes; during imaging, the frequency dimension of the single-channel echo signal is first converted into an intuitive distance, avoiding the problem of large computational complexity caused by the frequency dimension superposition operation; the spatial and temporal dimension information of multiple frames of images is integrated, and the construction process of the spatiotemporal feature model is derived; the spatiotemporal feature model is constructed to construct forward and backward spatiotemporal sensing factors, avoiding the occurrence of large-scale matrix calculations in the optimization process; the sparse distribution characteristics of the spatial position and respiratory frequency of the human target are utilized to obtain high signal-to-noise ratio spatiotemporal features through optimization.

[0119] In a second aspect, the present invention provides an adjacent hidden human target detection device based on time-space fusion, comprising:

[0120] A data preprocessing unit is used to process the target echo collected by the radar using non-causal moving target indication, remove the stationary target echo, and obtain an effective multi-channel echo signal;

[0121] An image acquisition unit, used for performing coherent imaging on effective multi-channel echo signals to acquire multiple frames of images;

[0122] A model building unit, used to build a preliminary spatiotemporal feature model using the acquired multiple frames of images;

[0123] The spatiotemporal feature acquisition unit is used to optimize the preliminary spatiotemporal feature model and obtain spatiotemporal features with a high signal-to-noise ratio.

[0124] In a third aspect, the present invention provides an electronic device comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned adjacent hidden human target detection method based on spatiotemporal fusion.

[0125] In a fourth aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned adjacent hidden human target detection method based on spatiotemporal fusion.

[0126] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting adjacent hidden human targets based on spatiotemporal fusion, characterized in that: The method comprises: Step 1: Use non-causal moving target indication to process the target echo collected by the radar, remove the static target echo, and obtain an effective multi-channel echo signal; Step 2: performing coherent imaging on the effective multi-channel echo signals to obtain multiple frames of images; Step 3: Use the acquired multi-frame images to build a preliminary spatiotemporal feature model; Step 4: Construct forward and backward space-time sensing factors based on the preliminary space-time feature model, construct and solve the constrained optimization objective function, and obtain space-time features with high signal-to-noise ratio; among them, the forward space-time sensing factor and backward spatiotemporal sensing factors The expression is: （14） Among them, the intermediate parameter expressions involved are: （15） and, and Respectively represent expanding the matrix in the brackets to four dimensions and five dimensions to satisfy the Hadamard product Requirements, , The intermediate parameters and The conjugate of is the Fourier transform matrix, , is the Fourier transform matrix The transpose and conjugate transpose forms of ; is the comprehensive representation of all pixels, is the number of transmitting antennas, is the number of receiving antennas, is the distance index, To remove the 1st, 2nd, ..., Tth frame signal after the stationary target echo, , and Represent the transmit antenna index, receive antenna index and frequency index respectively; = , and are the starting frequency and frequency modulation slope of the transmitted FM continuous wave signal, is the sampling time interval, The signal from the transmitting antenna to the grid point The time delay of returning to the receiving antenna, the subscript pmn represents the time delay from the transmitting antenna to the grid point Then returning to the receiving antenna, p represents the subscript index of the grid point, and P represents the total number of grid points; The frequency index representing the slow time direction; The constrained optimization objective function is constructed as: （17） The matrix form of the augmented Lagrangian function is: （18） in, is the regularization parameter, is the auxiliary matrix, and is the Lagrange multiplier and its conjugate transpose, represents the trace calculation of the matrix, is the penalty parameter, represents the 1-norm operation, represents the 2-norm operation; The optimization problem of formula (18) is converted into three sub-optimization problems. After derivation, the iterative solutions of the three sub-optimization problems are: （19） in, is the index of the iteration number, It is a soft threshold function, and the optimization result is obtained by iterative solution.

2. According to the method for detecting adjacent hidden human targets based on spatiotemporal fusion according to claim 1, it is characterized in that: The step 1 comprises: （1） in, Representative Frame original echo signal, Indicates Frame original echo signal, After removing the stationary target echo, Frame signal, Indicates the number of background frames. is the total number of frames after the stationary target echo is removed, Represents the original reflected echo signal, three dimensions , and Represent the transmit antenna index, receive antenna index and frequency index respectively.

3. The method for detecting adjacent hidden human targets based on spatiotemporal fusion according to claim 1, characterized in that: The step 2 comprises: Step 2.1, for the i-th frame signal after removing the stationary target echo, along the frequency index dimension, based on the fast inverse Fourier transform, the frequency dimension of the single-channel echo signal is converted into an intuitive distance representation; Step 2.2, the imaging area is evenly divided into rectangular grids, and the time delay of the single-channel echo signal from the transmitting antenna to the rectangular grid point and then back to the receiving antenna is calculated; Step 2.3, based on the time delay, calculating the coherent superposition of the echo signals of each channel, obtaining the pixel value of each rectangular grid point, and the pixel values ​​of each rectangular grid point are aggregated into the imaging result of the i-th frame signal; Step 2.4: Repeat steps 2.1 to 2.3 for each frame of the signal after removing the stationary target echo to obtain multiple frames of images.

4. The method for detecting adjacent hidden human targets based on spatiotemporal fusion according to claim 3, characterized in that: The step 3 comprises: Step 3.1, expanding the pixel value of each rectangular grid point along the time dimension; Step 3.2: Perform spectral analysis on the pixel values ​​of all expanded rectangular grid points to obtain a comprehensive representation of all pixel points and complete the construction of a preliminary spatiotemporal feature model.

5. The method for detecting adjacent hidden human targets based on spatiotemporal fusion according to claim 4, characterized in that: The step 3.1 includes, grid points The pixel value is expanded along the time dimension as follows: （8） In the formula, is the reflection coefficient of the human target, , and Represent the maximum values ​​of the transmit antenna index, receive antenna index and frequency index respectively. represents the slow time dimension, represents the center frequency, represents the ups and downs of the chest wall, K is the maximum value of the distance index, Represents the propagation speed of electromagnetic waves; The rise and fall of the chest wall Modeled as a sinusoidal wave: （9） in, Represents the amplitude of chest wall rise and fall, Represents the frequency of chest wall rise and fall.

6. The method for detecting adjacent hidden human targets based on spatiotemporal fusion according to claim 4, characterized in that: In step 3.2, The spectrum analysis is expressed as: （11） in, Representatives along Dimensions are subjected to a Fast Fourier Transform (FFT), represents the frequency index in the slow time direction, represents the Dirac function, represent Bessel series of the first kind; The processing of grid points is transformed into Expanded to a comprehensive representation of all pixels: （12） （13） in, Representative Frame sub-image, represents the matrix transpose operation, Indicates the starting frequency of transmitting FM continuous wave signal, Represents the grid point reached by the transmitting antenna The time delay back to the receiving antenna is , The frequency modulation slope of the transmitted FM continuous wave signal, is the sampling time interval, is the total number of frames after the stationary target echo is removed. middle, Dimensions provide spatial information. Slow time information is provided.

7. A device for detecting adjacent hidden human targets based on spatiotemporal fusion, applied to the method according to any one of claims 1 to 6, characterized in that: include: A data preprocessing unit is used to process the target echo collected by the radar using non-causal moving target indication, remove the stationary target echo, and obtain an effective multi-channel echo signal; An image acquisition unit, used for performing coherent imaging on effective multi-channel echo signals to acquire multiple frames of images; A model building unit, used to build a preliminary spatiotemporal feature model using the acquired multiple frames of images; The spatiotemporal feature acquisition unit is used to construct forward spatiotemporal sensing factors and backward spatiotemporal sensing factors based on the preliminary spatiotemporal feature model, construct and solve the constrained optimization objective function, and obtain spatiotemporal features with a high signal-to-noise ratio.

8. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; Wherein, when one or more programs are executed by the one or more processors, the one or more processors implement the adjacent hidden human target detection method based on spatiotemporal fusion as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: Executable instructions are stored thereon, which, when executed by a processor, enable the processor to implement the adjacent hidden human target detection method based on time-space fusion as described in any one of claims 1-6.

Images