A method and control device for reconstructing the geometry of an environment based on phaseless channel characteristics

Abstract

A method and a control device for environment geometry reconstruction based on phaseless channel features, the method comprising: obtaining phaseless channel features corresponding to multiple observations; extracting multipath components for the phaseless channel features of each observation; obtaining a distance difference measure of geometric distance and propagation distance in each observation according to geometric configuration information and the multipath components; performing an aggregation operation on each voxel based on a soft distance matching kernel and a power weight according to the multipath components and the distance difference measure to obtain an aggregation score corresponding to each voxel; obtaining an angle support range corresponding to each voxel according to the geometric configuration information, the distance difference measure and a preset truncation radius; and performing voxel occupancy determination and geometry reconstruction according to the aggregation score and the angle support range to obtain a binary geometry reconstruction result. The method solves the problems of phase dependence, improper aggregation operation unit, lack of distance error tolerance and inability to distinguish between real scattering and artifacts in the prior art.

Description

Technical Field

[0001] This application relates to the field of wireless signal processing technology, specifically to an environmental geometry reconstruction method and control device based on phaseless channel characteristics. Background Technology

[0002] During propagation, wireless signals are reflected and scattered by surrounding objects, forming multipath components (MPCs). The receiver can obtain channel characteristics describing the multipath structure of the wireless channel by processing the received signals. A typical form of this is the Channel Impulse Response (CIR), which characterizes the propagation delay and intensity of each multipath component in the time delay domain and naturally carries geometric information about the surrounding scatterers. Reconstructing the geometric structure of the environment from wireless channel characteristics provides crucial prior environmental information for applications such as environmental modeling, target recognition, and scene understanding, making it a valuable technical task.

[0003] Existing methods for reconstructing environmental geometry from wireless signals mainly fall into two categories. One category is coherent imaging methods based on complex-valued wavefield responses, such as synthetic aperture radar imaging and electromagnetic inverse scattering reconstruction, which require simultaneous use of the amplitude and phase of the received signal for imaging. However, in practical wireless channel measurement systems, due to limitations in hardware cost and distributed deployment, the system is prone to local oscillator drift, sampling timing deviations, and a lack of high-precision cross-viewpoint phase synchronization between different observation locations. Furthermore, dynamic changes in target boundary materials and observation angles can introduce unmeasurable reflection phase shifts. This makes it difficult to guarantee absolute phase reference in channel characteristic measurements, resulting in coherent imaging methods being unreliable in real-world scenarios.

[0004] Another type is the phase-free method based on amplitude information, such as incoherent backprojection. This method only utilizes amplitude-related information of the received signal (such as signal amplitude or power delay profile (PDP)) to backproject it into space according to geometric relationships and perform aggregation operations, thus avoiding dependence on phase in principle. However, this method has two significant drawbacks: First, it involves waveform cascading, with the entire amplitude response (including sidelobes and noise floor) participating in the spatial aggregation operation, resulting in a blurry and unclear edge in the final environmental reconstruction. Second, this method is a heuristic projection that does not explicitly model the actual observation process, failing to consider the distance estimation error caused by limited bandwidth, the physical migration characteristics of strong scattering points with changing viewing angles, and the spatial continuity of the contour structure in the angular domain. This makes the multi-view aggregation operation results highly susceptible to interference from accidental intersection artifacts, limiting the reliability of the reconstruction.

[0005] Therefore, how to fully exploit robust geometric constraints in multi-view channel feature observations without relying on phase information, eliminate spatial artifacts, and achieve low-complexity, high-reliability environmental geometry reconstruction is a key problem that urgently needs to be solved in the field of wireless signal environmental perception. Summary of the Invention

[0006] At least one embodiment of this application provides an environmental geometry reconstruction method and control device based on phaseless channel characteristics, which solves the problem of limited reliability in environmental geometry reconstruction in the prior art.

[0007] To solve the above-mentioned technical problems, this application is implemented as follows:

[0008] One embodiment of this application provides an environmental geometry reconstruction method based on phaseless channel characteristics, including:

[0009] The phase-free channel features corresponding to multiple observations are obtained. The phase-free channel features are channel features that do not depend on phase information and can reflect the received signal strength at each propagation delay.

[0010] Multipath components are extracted from the phaseless channel features of each observation to obtain the multipath components corresponding to each observation. The multipath components include time delay and power.

[0011] Based on the geometric configuration information and the multipath components, the distance difference metric between the geometric distance corresponding to each voxel in the voxel grid and the propagation distance corresponding to each multipath component is obtained in each observation. The voxel grid is obtained by discretizing a preset region of interest.

[0012] Based on the multipath components and the distance difference metric, an aggregation operation based on a soft distance matching kernel and power weight is performed on each voxel to obtain the aggregation score corresponding to each voxel.

[0013] Based on the geometric configuration information, the distance difference metric, and the preset cutoff radius, the angle support range corresponding to each voxel is obtained;

[0014] Based on the aggregation score and the angle support range, voxel occupancy determination and geometric reconstruction are performed to obtain the binary geometric reconstruction result.

[0015] Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the phaseless channel characteristics include any one of the following: power delay spectrum, channel impulse response amplitude, and logarithmic power.

[0016] Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the extraction methods for multipath components include at least one of the following: peak detection, constant false alarm rate (CFAR) detection, subspace methods (multiple signal classification (MUSIC), estimation of signal parameters via rotational invariance techniques (ESPRIT)), sparse recovery (orthogonal matching pursuit (OMP), least absolute shrinkage and selection operator (Lasso)), and deep learning).

[0017] Preferably, the environmental geometry reconstruction method based on phaseless channel features as described above, in scenarios where the phaseless channel feature is a power delay spectrum and the target location is known or estimable prior to the extraction of the multipath component, further includes:

[0018] Obtain the expected propagation distance range of the signal after being scattered by the target;

[0019] Based on the expected propagation distance range, a time delay window is obtained by performing a spatiotemporal conversion.

[0020] The power delay spectrum is filtered according to the delay window to obtain the power delay spectrum used for multipath component extraction.

[0021] Specifically, in the environmental geometry reconstruction method based on phase-free channel characteristics as described above, obtaining the expected propagation distance range of the signal scattered by the target includes:

[0022] The expected propagation distance range is determined based on the target location and transceiver location, target tracking, motion model prediction, or pre-set methods.

[0023] Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of obtaining the distance difference metric between the geometric distance corresponding to each voxel in the voxel grid and the propagation distance corresponding to each multipath component in each observation, based on the geometric configuration information and the multipath components, includes:

[0024] Based on the geometric configuration information, the sum of the distances from each voxel to the transmitter and receiver in each observation is determined as the geometric distance;

[0025] Based on the time delay corresponding to each of the multipath components in each observation, the propagation distance corresponding to each of the multipath components is obtained;

[0026] The difference between the geometric distance and the propagation distance in each observation is defined as the distance difference metric.

[0027] Alternatively, the distance difference metric can be determined as the ratio of the difference between the geometric distance and the propagation distance in each observation to the propagation distance.

[0028] Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of performing aggregation operations on each voxel based on a soft range matching kernel and power weights according to the multipath components and the distance difference metric to obtain the aggregation score corresponding to each voxel includes:

[0029] The soft distance matching kernel is obtained based on the distance difference metric and the preset cutoff radius in each observation. The soft distance matching kernel is a non-negative function that decreases monotonically as the absolute value of the distance difference metric increases, and is zero when the absolute value of the distance difference metric exceeds the preset cutoff radius.

[0030] According to the preset weight acquisition method, the power weight corresponding to each multipath component is obtained;

[0031] Based on the soft distance matching kernel and the power weight, the multipath components are aggregated to obtain the aggregated score corresponding to each voxel.

[0032] Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the soft range matching kernel is any one of the following: truncated Gaussian kernel, triangular kernel, double exponential kernel, inverse proportional kernel, and Epanechnikov kernel.

[0033] Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the preset weight acquisition method is any one of in-view power normalization, global power normalization, logarithmic weighting, square root weighting, and threshold weighting.

[0034] Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the aggregation operation includes: weighted summation, weighted averaging, summation in the logarithmic domain, or any other operation that can aggregate multiple multipath components.

[0035] Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of obtaining the angle support range corresponding to each voxel based on the geometric configuration information, the distance difference metric, and the preset truncation radius includes:

[0036] Based on the distance difference metric and the preset cutoff radius in each observation, the observation support status corresponding to each voxel is obtained. If there is a multipath component whose absolute value of the distance difference metric is less than or equal to the preset cutoff radius, the observation support status is determined to be supported; otherwise, the observation support status is determined to be unsupported.

[0037] Based on the single-station configuration or dual-station configuration indication and location information in the geometric configuration information, the observation direction corresponding to each voxel in each observation is determined. If the observation support status corresponding to the voxel is supported in any observation, the observation direction of the voxel in that observation is stored in the observation direction set corresponding to the voxel.

[0038] Based on the set of observation directions, the angular support range corresponding to each voxel is determined.

[0039] Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the angle support range is any one of the following: the maximum solid angle covered by the continuous support region on a unit sphere, the solid angle spanned by the convex hull of the observation direction, the total solid angle of the supported block, the proportion of the supported viewing angle, and the width of the continuous support interval after projection onto the main motion plane. The continuous support region is the region where the observation direction is continuously adjacent in the direction.

[0040] Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of performing voxel occupancy determination and geometry reconstruction based on the aggregation score and the angle support range to obtain a binary geometry reconstruction result includes:

[0041] Obtain the aggregate score threshold and the angle support range threshold;

[0042] The score determination result is determined based on the aggregated score and the aggregated score threshold.

[0043] The angle determination result is determined based on the angle support range and the angle support range threshold.

[0044] Based on the score determination result, the angle determination result, and the preset joint determination method, the binary occupancy indicator corresponding to the target voxel is determined. When the binary occupancy indicator corresponding to the target voxel is a first value, it indicates that the target voxel is occupied. When the binary occupancy indicator corresponding to the target voxel is a second value, it indicates that the target voxel is not occupied. The target voxel can be any one of the voxels.

[0045] Geometric reconstruction is performed based on the binary occupancy indicators corresponding to all voxels to obtain the binary geometric reconstruction result.

[0046] Specifically, in the environmental geometry reconstruction method based on phase-free channel characteristics as described above, obtaining the aggregation score threshold includes:

[0047] Based on the aggregation scores, the data are sorted from largest to smallest, and the critical cumulative scores of the voxels at the first preset ratio are determined as the aggregation score threshold.

[0048] Alternatively, a pre-set absolute threshold can be obtained as the aggregation score threshold;

[0049] Alternatively, the aggregated score threshold can be adaptively determined based on a preset table relating the score threshold to motion configuration, geometric configuration, and / or quantity configuration.

[0050] Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the angle support range threshold is one of the following: a preset coverage solid angle value, a minimum number of consecutive supported viewing angles, and an angle coverage rate.

[0051] Specifically, the environmental geometry reconstruction method based on phaseless channel characteristics, as described above, uses one of the following joint determination methods: logical AND, weighted summation, product rule, or other determination methods that simultaneously reflect the aggregation score and angular persistence.

[0052] Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of performing voxel occupancy determination and geometry reconstruction based on the aggregation score and the angle support range to obtain a binary geometry reconstruction result includes:

[0053] The aggregation score and the angle support range are input into the pre-trained voxel occupancy determination model to obtain the binary occupancy indicator or occupancy probability corresponding to each voxel output by the voxel occupancy determination model. The voxel occupancy determination model is a model that is pre-trained based on historical data and jointly determines the binary occupancy indicator or occupancy probability corresponding to the voxel based on the aggregation score and the angle support range.

[0054] The binary geometric reconstruction result is obtained by reconstructing based on the binary occupancy indicator or occupancy probability corresponding to all the voxels.

[0055] Preferably, the environmental geometry reconstruction method based on phaseless channel characteristics as described above further includes:

[0056] The rendering output is based on the binary geometric reconstruction result, wherein the rendering output is at least one of geometric point cloud, 3D point cloud, occupied grid, isosurface and triangular mesh.

[0057] Another embodiment of this application provides a control device for environmental geometry reconstruction based on phaseless channel characteristics, comprising:

[0058] The first processing module is used to obtain phase-free channel features corresponding to multiple observations. The phase-free channel features are channel features that do not depend on phase information and can reflect the received signal strength at each propagation delay.

[0059] The second processing module is used to extract multipath components from the phaseless channel features of each observation to obtain the multipath components corresponding to each observation. The multipath components include time delay and power.

[0060] The third processing module is used to obtain the distance difference metric between the geometric distance of each voxel in the voxel grid and the propagation distance of each multipath component in each observation, based on the geometric configuration information and the multipath components. The voxel grid is obtained by discretizing a preset region of interest.

[0061] The fourth processing module is used to perform aggregation operations on each voxel based on a soft distance matching kernel and power weight according to the multipath components and the distance difference metric, so as to obtain the aggregation score corresponding to each voxel;

[0062] The fifth processing module is used to obtain the angle support range corresponding to each voxel based on the geometric configuration information, the distance difference metric, and the preset cutoff radius.

[0063] The sixth processing module is used to perform voxel occupancy determination and geometric reconstruction based on the aggregation score and the angle support range, and obtain the binary geometric reconstruction result.

[0064] Another embodiment of this application provides an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described above.

[0065] Another embodiment of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described above.

[0066] Another embodiment of this application provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described above. Compared with the prior art, the environmental geometry reconstruction method and control device based on phaseless channel characteristics provided by this application, based on the joint processing of phaseless channel characteristics, soft range matching kernel, power weight, and angle support range, solves the problems of phase dependence, inappropriate aggregation operation unit, lack of range error tolerance, and inability to distinguish between real scattering and artifacts in existing back projection methods. Moreover, the overall computational complexity of the technical solution of this application is linearly related to the number of observations, the number of voxels, and the number of multipath components, without the need for a complex nonlinear iterative optimization process; it does not require additional dedicated hardware or dedicated waveforms, saving system costs and facilitating real-time deployment on general-purpose computing platforms. At the same time, it does not limit the motion state, trajectory shape, geometric configuration, and number configuration of the transceiver and target, and is applicable to various geometric and number configurations, thus having wide applicability. Attached Figure Description

[0067] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0068] Figure 1 This is one of the flowcharts illustrating the environmental geometry reconstruction method based on phaseless channel characteristics proposed in this application;

[0069] Figure 2 This is the second flowchart illustrating the environmental geometry reconstruction method based on phaseless channel characteristics proposed in this application.

[0070] Figure 3 This is the third flowchart illustrating the environmental geometry reconstruction method based on phaseless channel characteristics proposed in this application.

[0071] Figure 4 This is the fourth flowchart illustrating the environmental geometry reconstruction method based on phaseless channel characteristics proposed in this application.

[0072] Figure 5 This is the fifth flowchart illustrating the environmental geometry reconstruction method based on phaseless channel characteristics proposed in this application.

[0073] Figure 6 This is the sixth flowchart illustrating the environmental geometry reconstruction method based on phaseless channel characteristics proposed in this application.

[0074] Figure 7This is the seventh flowchart illustrating the environmental geometry reconstruction method based on phaseless channel characteristics proposed in this application.

[0075] Figure 8 A schematic diagram of a drone inspecting a building.

[0076] Figure 9 A schematic diagram illustrating how roadside base stations can detect passing vehicles;

[0077] Figure 10 A diagram illustrating a person holding a mobile phone walking in an indoor corridor to reconstruct their surroundings;

[0078] Figure 11 A schematic diagram illustrating the collaborative sensing of a moving human body by multiple dual-station transceivers indoors;

[0079] Figure 12 This is a schematic diagram of the control device for environmental geometry reconstruction based on phaseless channel characteristics according to this application. Detailed Implementation

[0080] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0081] The terms “first,” “second,” etc., used in the specification and claims of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The terms “and / or” in the specification and claims indicate at least one of the connected objects.

[0082] It should be noted that this application does not impose any restrictions on the motion state of the transceiver and the environment / target, the shape of the trajectory, the relative positional relationship between the region of interest and the observation trajectory, or the geometric configuration and number of transceivers themselves.

[0083] Specifically, this application applies to any of the following motion configurations:

[0084] (A) Static Environment / Target + Mobile Transceiver: The transceiver moves along an arbitrary known trajectory, acquiring channel impulse responses to a static environment or target at multiple locations. The region of interest can be located inside or outside the transceiver's trajectory, intersect with the trajectory, or surround the entire trajectory. This application does not limit the relative positional relationship between the region of interest and the transceiver's trajectory. Typical application scenarios include, but are not limited to: (a) transceiver observing a single external target to reconstruct the target's geometric contour (e.g., UAV inspection around an object); (b) transceiver moving within a scene to reconstruct the geometry of the surrounding environment (e.g., indoor mobile device scanning to reconstruct room layout, augmented reality (AR) / virtual reality (VR) device mapping, robot motion mapping in the environment); (c) transceiver moving along the side of a target to reconstruct the visible side contour of the target.

[0085] (B) Moving target + stationary transceiver: The transceiver is stationary (e.g., a fixed transmitter-receiver node), and the target moves along a real physical path (e.g., a vehicle drives by, a pedestrian crosses the road, or a mobile robot moves in the scene). Channel impulse responses are collected at predetermined time intervals during the target's movement.

[0086] (C) Moving target + moving transceiver: The transceiver and the target move simultaneously. As long as their relative positions are known at each observation, this configuration is included.

[0087] (D) Stationary target + stationary transceiver: This is only meaningful in a multi-transceiver system, where each transceiver is stationary and deployed in different locations to collect and observe the same stationary target or environment.

[0088] Furthermore, this application applies to any of the following geometric configurations:

[0089] (I) Single-site configuration: The transmitter and receiver are located in the same location;

[0090] (II) Dual-station configuration: The transmitter and receiver are located in different locations.

[0091] and any of the following quantity configurations:

[0092] (i) Single transceiver system: The system has only one transmitter-receiver pair, providing multiple perspectives through multiple observations;

[0093] (ii) Multi-transceiver system: The system has multiple transmitters and / or multiple receivers. Each pair of transceivers can independently or simultaneously collect observations, and all observations are incorporated into a unified aggregation calculation process.

[0094] The above-mentioned motion configuration, geometric configuration and quantity configuration can be combined arbitrarily, and this application does not restrict the combination method.

[0095] It should be noted that the geometric configuration information mentioned in this application refers to information related to the spatial geometry of the transceiver and voxels, including the position of the transmitter and receiver in each observation, an indication of whether the observation is a single-station or dual-station configuration, the spatial extent of the region of interest, and the center position of each voxel in the voxel grid.

[0096] It should also be noted that the observation configuration refers to the configuration information related to the observation process itself, including the total number of effective observations M, the sequence number of the acquisition time or acquisition location of each observation, and the identifier of the transmitter-receiver link corresponding to each observation.

[0097] To uniformly describe the observation process under various configurations, a "single effective observation" is defined as a channel impulse response measurement completed by a pair of transmitters and receivers at a certain moment. Assume that the method described in this application collects a total of M effective observations. Under different configurations, the M observations can originate from: (a) continuous measurements of a single transceiver system at M different moments (applicable to motion configurations A, B, and C); (b) parallel measurements of M different transmitter-receiver pairs at the same moment in a multi-transceiver system (applicable to motion configuration D); (c) combined measurements of multiple pairs of transmitter-receiver pairs at multiple moments in a multi-transceiver system (applicable to motion and multi-node combined configurations).

[0098] Record No. The transmitter location corresponding to this observation is The receiver location is The channel impulse response is In single-site configuration Under dual-station configuration During each observation, , The location of the target is known or can be estimated; the location information can be obtained by any of the following methods: transceiver self-localization, target tracking module, external positioning system or estimation based on motion model, and this application does not limit the method of acquisition.

[0099] Please see Figure 1 One embodiment of this application provides an environmental geometry reconstruction method based on phaseless channel characteristics, comprising:

[0100] Step S101: Obtain the phase-free channel features corresponding to multiple observations. The phase-free channel features are channel features that do not depend on phase information and can reflect the received signal strength at each propagation delay.

[0101] Step S102: Extract multipath components from the phaseless channel features of each observation to obtain the multipath components corresponding to each observation. The multipath components include time delay and power.

[0102] Step S103: Based on the geometric configuration information and the multipath components, obtain the distance difference metric between the geometric distance corresponding to each voxel in the voxel grid and the propagation distance corresponding to each multipath component in each observation. The voxel grid is obtained by discretizing a preset region of interest.

[0103] Step S104: Based on the multipath components and the distance difference metric, perform aggregation operation on each voxel based on soft distance matching kernel and power weight to obtain the aggregation score corresponding to each voxel;

[0104] Step S105: Based on the geometric configuration information, the distance difference metric, and the preset cutoff radius, obtain the angle support range corresponding to each voxel;

[0105] Step S106: Based on the aggregation score and the angle support range, voxel occupancy determination and geometric reconstruction are performed to obtain the binary geometric reconstruction result.

[0106] In this embodiment, environmental geometry is reconstructed by acquiring the phase-free channel features corresponding to multiple observations (effective observations). Based on the characteristics that the phase-free channel features can reflect the received signal strength at each propagation delay and do not depend on phase information, the dependence of geometric reconstruction on phase information can be effectively avoided. This means that there is no need to share any absolute phase reference between observations, thus fundamentally avoiding the problem of inconsistent phase references between independent multi-view observations, such as hardware phase uncertainty and view-dependent phase introduced by the target material. It eliminates local oscillator drift, sampling timing deviation and cross-view phase desynchronization at the hardware level, as well as view-dependent reflection phase introduced by the target material and incident geometry at the physical level. This solves the problem that the coherent back projection method is difficult to apply in independent multi-view scenarios.

[0107] Then, multipath components are extracted from the phaseless channel features of each observation to obtain the multipath components corresponding to each observation, which serve as the basis for subsequent calculations. The multipath components include time delay and power. That is, this application only uses the time delay of the multipath components extracted from the phaseless channel features or the propagation distance and power obtained based on the time delay as the basic observation, completely avoiding the dependence on phase information.

[0108] Furthermore, the region of interest is discretized in advance to obtain a voxel grid. Based on the preset geometric configuration information and the multipath component, the distance difference metric between the geometric distance of each voxel in the voxel grid and the propagation distance of each multipath component is obtained in each observation. This distance difference metric can effectively characterize the degree of spatial mismatch between the voxel and the actual signal path.

[0109] Furthermore, based on the multipath components and distance difference metrics obtained above, aggregation operations based on soft range matching kernels and power weights are performed on each voxel to obtain the aggregation score corresponding to each voxel. Here, multipath components are used as the basic aggregation operation unit, rather than directly back-projecting the entire amplitude-type channel features. This effectively avoids the problems of sidelobes and noise floor involved in the aggregation operation caused by directly using the entire amplitude-type channel features, allowing the aggregation operation to focus on physically meaningful scattering events, which is beneficial to improving the accuracy of the reconstruction results. By introducing a soft range matching kernel function, observations with slight distance deviations but still physically correlated can participate in the aggregation operation in a continuously attenuated manner. This provides an explicit tolerance mechanism for distance estimation errors introduced by limited bandwidth and noise, as well as for the migration of target scattering points with the observation viewpoint, improving the effectiveness of multi-view aggregation operations and the stability of the reconstruction results. Furthermore, by introducing a power weight based on the received power for each multipath component, reliable observations with strong signal-to-noise ratios contribute significantly to the aggregation operation, while unreliable observations with weak signal-to-noise ratios contribute less. This reflects the reliability differences of different multipath components during the aggregation operation, which is beneficial for improving the robustness of the aggregation results to noise and unreliable observations. In addition, during the above aggregation operation, the angular distribution of each voxel supported by each observation viewpoint is preserved.

[0110] Then, based on geometric configuration information, distance difference metric, and preset cutoff radius, the angular support range corresponding to each voxel is obtained to characterize the observational support of each voxel. In this step, the angular support range is not only related to the geometric position of the transceiver, but also to the specific observation data. It can be filtered based on the observation data during the determination of the angular support range, so as to effectively distinguish between the aggregation operation generated by real scattering and the artifacts generated by different scattering structures accidentally having the same propagation distance.

[0111] Finally, based on the aggregation score and the angle support range, voxel occupancy determination is performed, and geometric reconstruction is carried out to obtain binary geometric reconstruction results. By utilizing the continuity of the real scattering structure in terms of angle, the aggregation operation generated by real scattering and the artifact aggregation operation generated by the accidental same propagation distance of different scattering structures can be effectively distinguished. While preserving the real geometric structure, artifacts are suppressed, which is conducive to improving the reconstruction quality.

[0112] In summary, this application, based on the joint processing of phase-free channel characteristics, soft range matching kernel, power weights, and angle support range, solves the problems of phase dependence, inappropriate aggregation operation units, lack of range error tolerance, and inability to distinguish between real scattering and artifacts in existing back-projection methods. Furthermore, the overall computational complexity of this application's technical solution is linearly related to the number of observations, voxels, and multipath components, eliminating the need for complex nonlinear iterative optimization processes; it also eliminates the need for additional dedicated hardware or waveforms, saving system costs and facilitating real-time deployment on general-purpose computing platforms. Simultaneously, it imposes no restrictions on the motion state, trajectory shape, geometric configuration, or number of transceivers and targets, making it applicable to various geometric and numerical configurations and possessing broad applicability.

[0113] It should be noted that when discretizing the region of interest to obtain a voxel mesh, the region of interest is discretized into... Voxels form a voxel grid. The center position is denoted as This application does not limit the specific form of the voxel mesh. The shape of the voxel can be a cube or any other shape that is easy to discretize; the distribution of the voxel mesh can be a regular uniform mesh, a non-uniform mesh, or an adaptive resolution mesh; it should also be noted that in a two-dimensional scene, the voxel degenerates into a planar mesh element.

[0114] In one specific embodiment, the phase-free channel feature includes, but is not limited to: power delay spectrum, amplitude of channel impulse response, logarithmic power of channel impulse response, and any other channel feature that does not depend on phase information and can reflect the received signal strength at each propagation delay. To facilitate understanding by those skilled in the art, the following uses power delay spectrum as an example to illustrate its acquisition steps.

[0115] First of all, The channel impulse response (CIR) was acquired in each observation. The channel impulse response can be obtained by performing correlation operations between the received signal and a known transmitted signal (e.g., pilot signal, dedicated probe signal, pseudo-random sequence, or linear frequency modulated signal).

[0116] Then, corresponding to each observation Calculate its power delay spectrum:

[0117]

[0118] in, For the first Channel impulse response of the second observation To delay the transmission time, This represents the squared amplitude. The power delay spectrum depends only on the amplitude information of the channel impulse response and is independent of phase uncertainties such as local oscillator drift.

[0119] It should also be noted that this application does not limit the specific extraction method for multipath components. These methods include, but are not limited to, any one of: peak detection, constant false alarm rate detection, subspace methods, sparse recovery, and deep learning. In a specific embodiment, when the phase-free channel feature is a power delay spectrum and the multipath component extraction method is peak detection, the step of extracting multipath components from the phase-free channel feature for each observation to obtain the multipath components corresponding to each observation can be specifically expressed as: [The text abruptly ends here, likely due to an incomplete sentence or a missing section.] Peak detection is performed to extract multipath components. Let the first peak be... Detected in this observation The multipath component, the first The multipath components are denoted as ,in The time delay of this multipath component, This represents the power of the multipath component. The propagation distance corresponding to the multipath component can be obtained by converting the time delay.

[0120] See Figure 2 Preferably, the environmental geometry reconstruction method based on phaseless channel characteristics as described above, in scenarios where the phaseless channel characteristic is a power delay spectrum and the target location is known or estimable prior to the extraction of the multipath component, further includes:

[0121] Step S201: Obtain the expected propagation distance range of the signal scattered by the target;

[0122] Step S202: Perform spatiotemporal conversion based on the expected propagation distance range to obtain a time delay window;

[0123] Step S203: Filter the power delay spectrum according to the delay window to obtain the power delay spectrum used for multipath component extraction.

[0124] To suppress interference from background scattering on reconstruction, in scenarios where the phaseless channel characteristic is a power delay profile (PDP) and the target position is known or estimable prior (e.g., transceiver observing an external target in motion configuration A, transceiver sensing a moving target in motion configuration B), range gating can be performed on the power delay profile before extracting the multipath components. Specifically, for each observation... Obtain the expected propagation distance range of the signal scattered by the target. ,in, and A spatial range slightly wider than the target is obtained to ensure that all of the target's own scattering falls within the window. Then, this propagation distance window is converted into a time delay window. ,in To calculate the speed of light, the power delay spectrum is finally filtered based on the time delay window, retaining only the power delay spectrum. Within the time delay window, the portion outside the window is set to zero or not involved in subsequent processing. After the above range gating, the power delay spectrum segment corresponding to the target scattering event is retained. Multipath components generated by the background environment (such as distant walls and ground clutter) that differ significantly from the target range are excluded from subsequent aggregation operations, thereby improving the accuracy of target geometry reconstruction and reducing the computational load of subsequent processing.

[0125] In scenarios where the target location cannot be obtained prior (e.g., in motion configuration A, the transceiver moves within the scene to reconstruct the unknown surrounding environment), the distance gating step can be omitted, and multipath components can be extracted directly from the complete power delay spectrum.

[0126] Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, obtaining the expected propagation distance range of the signal scattered by the target includes: calculating based on the target position and transceiver position, determining the expected propagation distance range based on target tracking, motion model prediction, or a pre-set method. Taking the calculation based on the target position and transceiver position as an example, it specifically involves calculating based on the known or estimable target position and the transmitter position. and receiver position , and then calculate to determine.

[0127] See Figure 3 Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of obtaining the distance difference metric between the geometric distance corresponding to each voxel in the voxel grid and the propagation distance corresponding to each multipath component in each observation, based on the geometric configuration information and the multipath components, includes:

[0128] Step S301: Based on the geometric configuration information, determine the sum of the distances from each voxel to the transmitter and receiver in each observation as the geometric distance;

[0129] Step S302: Based on the time delay corresponding to each multipath component in each observation, obtain the propagation distance corresponding to each multipath component;

[0130] Step S303: Determine the difference between the geometric distance and the propagation distance in each observation as the distance difference metric;

[0131] Alternatively, in step S304, the ratio of the difference between the geometric distance and the propagation distance in each observation to the propagation distance is determined as the distance difference measure.

[0132] In this embodiment, when obtaining the distance difference metric between geometric distance and propagation distance, firstly for each voxel... and each observation Calculate geometric distance It is defined as the sum of the distance from the voxel to the transmitter and the distance from the voxel to the receiver, and can be specifically expressed as:

[0133] in This represents the Euclidean distance. In a single-station configuration, due to the receiver... and transmitter For the same transceiver ,Right now At this point, the geometric distance is twice the distance from the voxel to the transceiver, which reduces the above equation to: .

[0134] In addition, each observation Multipath components Corresponding propagation distance , representing the total propagation path length of the signal from the transmitter through the scattering point to the receiver, is obtained by converting the time delay of the multipath components, i.e.:

[0135]

[0136] in, At the speed of light, Indicates observation Multipath components The time delay.

[0137] It should also be noted that, geometrically, under all geometric configurations, each multipath component corresponds to a major axis equal to the transmitter and receiver as the two foci. The ellipsoidal surface constraint; under a single-station configuration, the ellipsoid degenerates into a circle centered at the transceiver with a radius equal to... Spherical constraint.

[0138] Then, by subtracting the propagation distance from the geometric distance in each observation, the required distance difference metric can be obtained, which can be specifically expressed as: .

[0139] Alternatively, after obtaining the difference between the geometric distance and the propagation distance in each observation, the ratio of this difference to the corresponding propagation distance, i.e., the relative distance difference, can be used as the distance difference metric, which can be specifically expressed as: .

[0140] In another embodiment, other metrics that can reflect the degree of matching between geometric distance and propagation distance are also used as the distance difference metric of this application.

[0141] See Figure 4 Specifically, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of performing aggregation operations on each voxel based on a soft range matching kernel and power weights according to the multipath components and the distance difference metric to obtain the aggregation score corresponding to each voxel includes:

[0142] Step S401: Based on the distance difference metric and the preset cutoff radius in each observation, obtain the soft distance matching kernel. The soft distance matching kernel is a non-negative function that monotonically decreases as the absolute value of the distance difference metric increases, and is zero when the absolute value of the distance difference metric exceeds the preset cutoff radius.

[0143] Step S402: Obtain the power weight corresponding to each multipath component according to the preset weight acquisition method;

[0144] Step S403: Based on the soft distance matching kernel and the power weight, perform aggregation operation on the multipath components to obtain the aggregation score corresponding to each voxel.

[0145] In this embodiment, when obtaining the aggregation score based on multipath components and distance difference metrics, priority is given to each voxel. and each observation Each multipath component The corresponding distance difference metric and the preset cutoff radius are used to calculate the soft distance matching kernel. Among them, soft distance matching kernel It is a non-negative function that measures distance difference. The absolute value increases and monotonically decreases, and the distance difference metric exceeds the preset cutoff radius. The value is zero. Specifically, soft distance matching kernels include, but are not limited to: any one of the following: truncated Gaussian kernel, triangular kernel, double exponential kernel, inverse proportional kernel, and Epanechnikov kernel; or any other kernel that monotonically decreases with respect to the distance difference metric. Symmetry, in A non-negative function that takes the maximum value at a given point.

[0146] In one specific embodiment, the soft distance matching kernel The preferred option is a truncated Gaussian kernel, which can be specifically expressed as:

[0147]

[0148] in, The kernel function scaling parameter is preferably chosen to be proportional to the system distance resolution. ,Right now , For signal bandwidth, It is a direct proportionality constant. Preset cutoff radius. Preferred selection ,in The maximum spatial size of the target It is a proportionality constant used to tolerate the migration of the scattering point with the viewing angle.

[0149] To facilitate understanding by those skilled in the art, the formulas for other forms of soft distance matching kernels are illustrated below:

[0150] Triangular nucleus:

[0151]

[0152] Epanechnikov core:

[0153]

[0154] Dual exponential core:

[0155]

[0156] Inverse proportional kernel:

[0157]

[0158] Simultaneously, the power weights corresponding to each multipath component will be obtained according to a preset weight acquisition method, wherein the power weights... To reflect the relative reliability of multipath components during aggregation, the preset weight acquisition method is any one of the following: in-view power normalization, global power normalization, logarithmic weighting, square root weighting, and threshold weighting, or other non-negative weight calculation methods based on multipath component power. In a specific embodiment, the preset weight acquisition method preferably adopts in-view power normalization, in which case the power weight can be specifically expressed as:

[0159]

[0160] in, This represents the power of the i-th multipath component in the m-th observation; This represents all observations in the m-th observation. The sum of the power of each multipath component Iterate from 1 to That is, the proportion of the power of the multipath component to the total power in this observation.

[0161] To facilitate understanding by those skilled in the art, the following provides a simplified illustration of some forms of power weighting:

[0162] Log-weighted:

[0163]

[0164] Square root weighted:

[0165]

[0166] Threshold weighting: This means selecting multipath components with power exceeding a threshold for use in aggregation calculations.

[0167] Global power normalization: This means normalizing the total power of all observations and multipath components.

[0168] Finally, all the observations and multipath components obtained above are aggregated to obtain voxels. Aggregate fraction Optionally, the aggregation operation includes: weighted summation, weighted average, summation in the logarithmic field, or any other operation that can aggregate multiple said multipath components.

[0169] In one specific embodiment, taking the aggregation operation as a weighted summation as an example, it can be specifically expressed as follows:

[0170]

[0171] Among them, the aggregation fraction The larger the value, the more voxels it represents. The more reliable the multipath components in the more observational perspectives, the greater the likelihood that the voxel is occupied by the scatterer.

[0172] It should be noted that the normalization factor 1 / M in the above aggregate score is only an example. This application does not limit the normalization method. The actual number of observations supported by each voxel can also be used as the normalization factor, or the sum of the support contributions of each observation can be used directly as the aggregate score without normalization.

[0173] It should be noted that if the distance estimation accuracy meets the preset conditions during the aforementioned process of obtaining the distance difference metric, a hard distance matching kernel based on zero and a fixed value can be used instead of a soft distance matching kernel. Specifically, the hard distance matching kernel takes a fixed value when the distance difference metric does not exceed a preset cutoff radius, and takes zero otherwise. In this case, the kernel function degenerates into a rectangular window.

[0174] See Figure 5Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of obtaining the angle support range corresponding to each voxel based on the geometric configuration information, the distance difference metric, and the preset truncation radius includes:

[0175] Step S501: Based on the distance difference metric and the preset cutoff radius in each observation, obtain the observation support status corresponding to each voxel. If there is a multipath component whose absolute value of the distance difference metric is less than or equal to the preset cutoff radius, then the observation support status is determined to be supported; otherwise, the observation support status is determined to be unsupported.

[0176] Step S502: Based on the single-station configuration or dual-station configuration indication and location information in the geometric configuration information, determine the observation direction corresponding to each voxel in each observation. If the observation support status corresponding to the voxel is supported in any observation, then store the observation direction of the voxel in that observation into the observation direction set corresponding to the voxel.

[0177] Step S503: Determine the angle support range corresponding to each voxel based on the set of observation directions.

[0178] In this embodiment, the steps for obtaining the angular support range corresponding to each voxel are illustrated. First, based on the distance difference metric and the preset cutoff radius (which has been specifically explained above and will not be repeated here), the observation support status corresponding to each voxel is obtained. If there is a multipath component whose absolute value of the distance difference metric is less than or equal to the preset cutoff radius, the observation support status is determined to be supported; otherwise, the observation support status is determined to be unsupported. For ease of description, a binary support indicator is used in this embodiment. This indicates whether each voxel is supported by observations in each observation; specifically, voxels... In the The binary support indicator under this observation can be expressed as:

[0179]

[0180] Furthermore, define voxels In the The observation direction in this observation is a unit direction vector. , characterizing the The second observation was based on voxels. The three-dimensional spatial direction seen from the perspective. Then, based on the single-station or dual-station configuration indication and position information in the geometric configuration information, the observation direction corresponding to all voxels in each observation is determined. Specifically, in the case of a single-station configuration indication in the geometric configuration information, it is preferably defined as the direction from the voxel. (Voxenon center location) Pointing to the transceiver Unit vector:

[0181]

[0182] When the geometric configuration information indicates a bistation configuration, it is preferable to define the observation direction as the bistation bisector direction, i.e., from the voxel. (Voxenon center location) Pointing to the transmitter The unit vector and the direction of the receiver The normalized direction of the sum of unit vectors, i.e.:

[0183]

[0184] in, , And in the receiver With the transmitter When locations overlap or are integrated, the configuration naturally degenerates into the definition described above for a single-site configuration.

[0185] Optionally, the observation direction can also adopt any direction that can characterize the first... The second observation was in voxels (Voxenon center location) The equivalent geometric direction quantity from the perspective of each observation direction. A point on the corresponding unit sphere.

[0186] It should also be noted that, in cases where the distance between the transceiver and the center of the region of interest is much larger than the spatial extent of the region of interest, to simplify calculations, the observation perspective of all voxels can be unified as the transceiver relative to the center of the region of interest. The direction, that is, to make all voxels share the same observation angle sequence. This application does not limit the specific definition of the observation direction; any direction that can characterize the first... The quantities of geometric directions observed in each observation from a voxel perspective are all within the scope of protection of this application.

[0187] Furthermore, for each voxel, based on its observation support status in each observation and the definition of the observation direction mentioned above, if the observation support status corresponding to the voxel is supportive in any observation, then the observation direction of the voxel in that observation is stored in the observation direction set corresponding to the voxel, i.e., the set of all supported observation directions. middle.

[0188] Then, the corresponding angle support range can be determined based on the set of observation directions and the specific type of angle support range. In one embodiment, the angle support range is any one of the following: the maximum solid angle covered by the continuous support area on a unit sphere, the solid angle spanned by the convex hull of the observation direction, the total solid angle of the supported block, the proportion of the supported viewing angle, and the width of the continuous support interval after projection onto the main motion plane. The continuous support area is the area where consecutively adjacent observation directions are located.

[0189] Taking the maximum solid angle covered by a continuous support region on a unit sphere as an example, when obtaining the angle support range, firstly, the supported directions of each voxel in the observation direction set that are continuously adjacent in direction (i.e., the angle between the directions is less than a preset proximity threshold) are grouped into the same continuous support region. Each supported direction in the continuous support region covers a certain solid angle on a unit sphere. Then, the maximum solid angle covered by each continuous support region on a unit sphere for the same voxel is taken as the angle support range corresponding to that voxel. ,Right now:

[0190]

[0191] in, For the first A subset of observation directions contained within a continuous support region. This represents the solid angles covered by these directions on a unit sphere. These solid angles can be calculated using the convex hull area of ​​the supported observation directions on the unit sphere, or by dividing the unit sphere into blocks and calculating the total solid angles of the occupied blocks.

[0192] It should be noted that in the special case where all observation directions are coplanar (i.e., all observation directions lie in the same plane, such as when a transceiver moves approximately in the same plane), the unit sphere degenerates into a unit circle, the solid angle degenerates into an angle interval on the circumference, and the continuous support region degenerates into a continuous support interval. In this case, the angle support range... It degenerates into the angle width of the maximum continuous support interval.

[0193] It should also be noted that the solid angle subtended by the convex hull of the above-mentioned observation direction is the solid angle subtended by the convex hull of the supported direction on the unit sphere (not required to be continuous).

[0194] The total solid angle of the supported blocks is calculated by dividing a unit sphere into several blocks and counting the total solid angle of the blocks occupied by the supported direction.

[0195] The percentage of supported viewpoints, which is the ratio of the number of supported observation directions to the total number of observation directions;

[0196] The width of the continuous support interval after projection onto the main motion plane is the width of the continuous support interval in the azimuth angle after the three-dimensional observation direction is projected onto the main motion plane.

[0197] In another embodiment, the angular support range can also be other measures that can reflect the breadth or continuity of the coverage of the observation direction on a unit sphere.

[0198] See Figure 6 Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of performing voxel occupancy determination and geometry reconstruction based on the aggregation score and the angle support range to obtain a binary geometry reconstruction result includes:

[0199] Step S601: Obtain the aggregation score threshold and the angle support range threshold;

[0200] Step S602: Determine the score determination result based on the aggregated score and the aggregated score threshold;

[0201] Step S603: Determine the angle determination result based on the angle support range and the angle support range threshold.

[0202] Step S604: Based on the score determination result, the angle determination result, and the preset joint determination method, determine the binary occupancy indicator corresponding to the target voxel. When the binary occupancy indicator corresponding to the target voxel is a first value, it indicates that the target voxel is occupied. When the binary occupancy indicator corresponding to the target voxel is a second value, it indicates that the target voxel is not occupied. The target voxel can be any one of the voxels.

[0203] Step S605: Perform geometric reconstruction based on the binary occupancy indicators corresponding to all voxels to obtain the binary geometric reconstruction result.

[0204] In this embodiment, when performing voxel occupancy determination and geometric reconstruction, the aggregation score threshold corresponding to the aggregation score and the angle support range threshold corresponding to the angle support range are first obtained for determination, and the corresponding score determination results and angle determination results are obtained respectively. Further, based on the above score determination results, angle determination results, and a preset joint determination method, the binary occupancy indicator corresponding to the target voxel can be jointly determined. When the binary occupancy indicator corresponding to the target voxel is a first value (e.g., 1), it indicates that the target voxel is occupied; when the binary occupancy indicator corresponding to the target voxel is a second value (e.g., 0), it indicates that the target voxel is not occupied. The voxel occupancy status can be determined based on this binary occupancy indicator. Finally, based on all binary occupancy indicators... The first value occupies the voxel (voxel center position) The set of ) That is, the corresponding target or environment, the corresponding geometric structure is reconstructed in the region of interest, which is the binary geometric reconstruction result.

[0205] In one specific embodiment, the step of obtaining the aggregation score threshold may include:

[0206] Based on the aggregation scores, they are sorted from largest to smallest, and the critical cumulative score of the voxels at the first preset proportion is determined as the aggregation score threshold, i.e., according to ( (Rules) determined;

[0207] Alternatively, obtain a pre-set absolute threshold ( The aggregation score threshold is denoted as .

[0208] Alternatively, the aggregated score threshold can be adaptively determined based on a preset table relating the score threshold to motion configuration, geometric configuration, and / or quantity configuration.

[0209] It should be noted that, based on this application, determining the aggregation score threshold according to other thresholding methods based on aggregation scores to achieve the corresponding technical effect should also fall within the protection scope of this application.

[0210] In another embodiment, the aforementioned angle support range threshold can be one of the following: a preset solid angle value, a minimum number of consecutive supported viewing angles, and an angle coverage rate. Preferably, the angle support range is determined during the acquisition process to ensure that the angle determination result can be determined based on the aforementioned angle support range and the angle support range threshold.

[0211] In another embodiment, the aforementioned joint determination method can be one of logical AND, weighted summation, product rule, or other determination methods that simultaneously reflect the aggregation score and angular persistence. Taking logical AND as an example, the determination steps can be expressed as follows: when the aggregation score corresponding to any voxel is greater than or equal to the aggregation score threshold, and the corresponding angular support range is greater than or equal to the angular support range threshold... If the binary occupancy indicator is set to the first value, then the binary occupancy indicator is set to the second value.

[0212] Then, these occupying voxels are arranged in space according to their central positions. Once marked or rendered, the geometric reconstruction results of the target or environment are obtained, which can be used for subsequent target recognition, environment modeling, digital twin channel construction or other applications.

[0213] It should be noted that the angle persistence threshold can also be achieved using the minimum number of consecutive supported viewpoints, angle coverage, or other angle-support-based thresholding methods. The combination of the amplitude threshold and the angle persistence threshold can also employ weighted summation, product rules, or other joint decision-making methods that simultaneously reflect amplitude and angle persistence.

[0214] See Figure 7 Preferably, in the environmental geometry reconstruction method based on phaseless channel characteristics as described above, the step of performing voxel occupancy determination and geometry reconstruction based on the aggregation score and the angle support range to obtain a binary geometry reconstruction result includes:

[0215] Step S701: Input the aggregate score and the angle support range into the pre-trained voxel occupancy determination model, and obtain the binary occupancy indicator or occupancy probability corresponding to each voxel output by the voxel occupancy determination model. The voxel occupancy determination model is a model that is pre-trained based on historical data and jointly determines the binary occupancy indicator or occupancy probability corresponding to the voxel based on the aggregate score and the angle support range.

[0216] Step S702: Reconstruct the binary geometric reconstruction result by performing reconstruction based on the binary occupancy indication or occupancy probability corresponding to all the voxels.

[0217] In this embodiment, another method for obtaining binary geometric reconstruction results by performing voxel occupancy determination and geometric reconstruction based on the aggregation score and the angle support range is also described. That is, the mapping from aggregation score to voxel occupancy determination, in addition to the thresholding and joint decision method based on preset rules, can also be implemented using a machine learning model. Specifically, historical data (including but not limited to aggregation scores) will be collected or simulated in advance. Angle support range and voxel occupancy label Then, pre-training is performed based on this historical data, including but not limited to collecting several sets of samples from the historical data, each set of samples including the aggregated scores obtained as described above. (and optional angle support range) ) and their corresponding voxel occupancy labels Then, the decision model is trained offline using the aggregate score (and optional angle support range) as input and the voxel occupancy label as supervision; after training, the required voxel occupancy decision model can be obtained.

[0218] Then, in the reconstruction phase (i.e., in practical applications), the aggregation score (and optional angle support range) obtained from the aforementioned steps is input into the voxel occupancy determination model. The model directly outputs the occupancy probability or binary occupancy indicator of each voxel. Furthermore, geometric reconstruction can be performed based on the occupancy probability or binary occupancy indicator to obtain the binary geometric reconstruction result. By learning the statistical differences between the true scattering structure and artifacts in terms of aggregation score (and angle support) in a data-driven manner, artifacts can be suppressed while preserving the true geometry.

[0219] It should be noted that when performing geometric reconstruction based on the occupancy probability, the occupancy probability is first thresholded. That is, the voxel is determined to be occupied only if the occupancy probability corresponding to a voxel is greater than the preset probability.

[0220] It should also be noted that the voxel occupancy determination model can be a multilayer perceptron, convolutional neural network, U-Net, or other network structure that can establish a mapping between aggregate scores and occupancy determination.

[0221] Preferably, the environmental geometry reconstruction method based on phaseless channel characteristics as described above further includes:

[0222] The rendering output is based on the binary geometric reconstruction result, wherein the rendering output is at least one of geometric point cloud, 3D point cloud, occupied grid, isosurface and triangular mesh.

[0223] In this embodiment, the rendering output will be based on the binary geometric reconstruction results obtained above. That is, for the voxel with the binary occupancy indicator value of the first value, the rendering will be performed according to its center position and output, so that the user can intuitively perceive the effect of environmental geometric reconstruction.

[0224] To facilitate understanding of the technical solutions of this application by those skilled in the art, specific embodiments in various application scenarios are given below based on at least one of the above solutions.

[0225] First Embodiment: This embodiment describes an application scenario of using a drone to reconstruct the external geometric contour of a stationary building. See also... Figure 8 , Figure 8 This is a schematic diagram of a drone inspecting a building. The drone, equipped with a transceiver, flies around a stationary building along a preset flight path, collecting channel impulse responses at multiple locations during the flight. The three-dimensional external contour of the building is reconstructed using the method described in this application. This method can be used for building surveying, facade inspection, and post-disaster building structure assessment. This embodiment corresponds to the sub-case of "region of interest located inside the transceiver trajectory" in motion configuration A described in this application. The geometric configuration is a single-station configuration, and the quantity configuration is a single transceiver system.

[0226] The system parameters used in this embodiment are as follows:

[0227] Transmitted signal: Orthogonal Frequency Division Multiplexing (OFDM) signal or linear frequency modulated signal, center frequency ,bandwidth Corresponding distance resolution ;

[0228] Observation configuration: A single UAV equipped with a transceiver flies around the building along a preset flight path. Channel impulse responses were collected at each observation location, and the flight position was jointly provided by the UAV's Global Positioning System (GPS) and Inertial Measurement Unit (IMU).

[0229] Geometric configuration: Single-site, meaning the transceiver's transmitting and receiving antennas are located in the same position on the UAV. ;

[0230] Target: A stationary building with a maximum horizontal dimension D of approximately The radius of the flight trajectory is approximately 2 to 3 times the size of the target.

[0231] Region of Interest: A rectangular area covering the building's perimeter, typically sized as follows. ;

[0232] Voxel mesh: 3D uniform mesh, voxel side length set to ,common Voxels (when computational resources are limited, the voxel side length can be appropriately increased to reduce the number of voxels);

[0233] Parameters of the soft distance matching kernel function: Cut-off radius ;

[0234] Threshold parameter: The aggregate score threshold is the aggregate score. Top voxels, angular support range threshold ;

[0235] Signal processing hardware: Onboard embedded processors, general-purpose central processing units (CPUs), graphics processing units (GPUs), or dedicated digital signal processors (DSPs) can be used to run back-end algorithms, or the collected data can be transmitted back to the ground station for processing.

[0236] The specific implementation steps are as follows:

[0237] Step 1: The UAV transmits pilot signals from 72 flight positions and receives the echo signals reflected by buildings. By performing correlation calculations between the received signals and the known transmitted pilot signals, 72 sets of channel impulse responses are obtained. For each Square the amplitude to obtain the corresponding power delay spectrum. .

[0238] Step 1A: Distance gating based on prior building distances. In this embodiment, the building locations are fixed and known, and the UAV's flight positions are jointly provided by the Global Positioning System and the Inertial Measurement Unit, thus distance gating can be performed. For each flight position... Based on the known building locations and transceiver locations Determine the expected round-trip propagation distance window of the signal scattered by the building. ,in and Obtain a spatial range slightly wider than the building, convert it into a time delay window, and retain only the power time delay spectrum. Within this window, the portion outside the window is set to zero. After this gating, subsequent multipath component extraction is performed only on the time delay segments belonging to the building, while clutter from other distant buildings or the ground is excluded.

[0239] Step 2: For each Peak detection is performed using a threshold-based or constant false alarm rate (CFAR) peak detection algorithm to extract local peaks in the power delay spectrum. Each peak corresponds to a multipath component generated by a scattering point on the building, and its delay is extracted. and power .

[0240] Step 3: Select the region of interest Evenly divided into Voxels, recording each voxel center coordinates .

[0241] Step 4: For each voxel (voxel center location) ) and each observation Calculate the geometric distance based on the single-station configuration. Calculate the propagation distance of the response for each multipath component. .

[0242] Step 5: For each voxel, for each observation Each multipath component Calculate the distance difference metric Calculate the truncated Gaussian kernel. Calculate power weights Voxels are obtained through aggregation operations. Aggregate fraction .

[0243] Step 6: Calculate the angular support range for each voxel. For each voxel, define its position on the first... The observation direction in this observation is a unit vector pointing from the voxel to the transceiver. When multipath components exist Make Seasonal binary support indicators Otherwise Examine the set of all supported observation directions for this voxel. Grouping consecutively adjacent supported directions into the same continuous support region, and calculating... This represents the maximum solid angle covered on a unit sphere across all continuously supported regions.

[0244] Step 7: Aggregation fraction for all voxels Sort and keep the first few. voxels; at the same time, it requires A voxel that satisfies both of these conditions is designated as occupied. The remaining voxels are set to unoccupied ( The final binary geometric reconstruction result is obtained. .

[0245] After the above steps, the occupied voxel The resulting structure is the geometric reconstruction of the building. This result can be used for applications such as building surveying, facade inspection, and rapid post-disaster structural assessment. This embodiment uses a drone inspecting a building as an example, but this application is also applicable to other scenarios where a single transceiver observes a stationary external target, such as a drone inspecting a wind turbine or transmission tower, or a robot observing equipment or shelves to reconstruct its outline.

[0246] The second embodiment describes an application scenario where a roadside stationary base station is used to reconstruct the geometric contours of passing vehicles. See also... Figure 9 , Figure 9This is a schematic diagram of a roadside base station sensing a passing vehicle. A transceiver is stationary on one side of the road (e.g., mounted on a lamppost, traffic signal pole, or roadside unit). As a vehicle passes by the base station, the base station continuously collects channel impulse responses at predetermined time intervals during the vehicle's movement. Because the vehicle's position changes over time, the relative position between the base station and the vehicle differs at different collection times, providing multi-view observation. The three-dimensional geometric contour of the vehicle can be reconstructed using the method described in this application, which can be used for vehicle type recognition, traffic flow monitoring, vehicle classification and tolling, etc. This embodiment corresponds to the motion configuration B (moving target + stationary transceiver) described in this application, with a single-site geometric configuration and a single transceiver system.

[0247] The following system parameters are used in this embodiment:

[0248] Transmitted signal: Native pilot signal of the communication system, center frequency ,bandwidth Corresponding distance resolution ;

[0249] Observation configuration: A single transceiver base station is stationary on one side of the road. Channel impulse responses are collected at fixed time intervals (e.g., every few milliseconds) as vehicles pass by. A total of [number] impulse responses are collected. The time of collection is recorded as follows: ;

[0250] Geometric configuration: Monostation, meaning the transmitting and receiving antennas are located in the same position, which remains constant throughout all data acquisition times, denoted as _____. ;

[0251] Objective: The maximum horizontal dimension D of a vehicle traveling along the road is approximately... Typical driving speed is ;

[0252] Target location acquisition: The vehicle's location at each data collection time is reported to the base station by its own positioning module.

[0253] Region of Interest: A cuboid region in a vehicle-mounted coordinate system, with the vehicle as the reference. Typical dimensions. ;

[0254] Voxel mesh: 3D uniform mesh, voxel side length set to ,common Individual factors;

[0255] Parameters of the soft distance matching kernel function: Cut-off radius ;

[0256] Threshold parameter: The aggregate score threshold is the aggregate score. Top voxels, angular support range threshold ;

[0257] Signal processing hardware: Embedded processors in roadside units, CPUs / GPUs in edge computing nodes can be used, or the collected data can be uploaded to a traffic management cloud platform for processing.

[0258] The specific implementation steps are as follows:

[0259] Step 1: The base station transmits pilot signals at 60 different times during the vehicle's passage and receives the echo signals reflected by the vehicle. By performing correlation calculations between the received signals and the known transmitted pilot signals, 60 sets of channel impulse responses are obtained. For each Square the amplitude to obtain the corresponding power delay spectrum. .

[0260] Step 1A: Distance gating based on vehicle prior distance. In this embodiment, the vehicle's position at each data collection time is provided by its own positioning module, thus distance gating can be performed. For each data collection time... Based on the known vehicle location and base station location T, determine the expected round-trip propagation distance window of the signal scattered by the vehicle. ,in and Obtain a spatial range slightly wider than the vehicle's, convert it into a time delay window, and retain only the power time delay spectrum. Within this window, the portion outside the window is set to zero. After this gating, subsequent multipath component extraction is performed only on the time delay segments belonging to the vehicle, while background scattering from roads, roadside buildings, other vehicles, etc., is excluded.

[0261] Step 2: For each Peak detection is performed using a threshold-based or constant false alarm rate (CFAR) peak detection algorithm to extract local peaks in the power delay spectrum. Each peak corresponds to a multipath component generated by a scattering point on the vehicle, and its delay is extracted. and power .

[0262] Step 3: Since the vehicle is in motion, to enable multi-view observations to be aggregated and calculated on a unified grid, a vehicle-mounted coordinate system is established with the vehicle as a reference. Using the vehicle's position at each acquisition time provided by its self-positioning module as described in the system parameters, the base station position is determined. Transform to the vehicle coordinate system to obtain the first... The equivalent transceiver position in the vehicle coordinate system was observed. The region of interest in the vehicle coordinate system. Evenly divided into Voxel, recording each voxel center coordinates .

[0263] Step 4: For each voxel (voxel center location) ) and each observation Calculate the geometric distance based on the single-station configuration. Calculate the propagation distance of the response for each multipath component. .

[0264] Step 5: For each voxel, for each observation Each multipath component Calculate the distance difference metric Calculate the truncated Gaussian kernel. Calculate power weights Aggregation operations yield voxels. Aggregate fraction .

[0265] Step 6: Calculate the angular support range for each voxel. For each voxel, define its position on the first... The observation direction in this observation is a unit vector pointing from the voxel to the transceiver. When multipath components exist Make season Otherwise Examine the set of all supported observation directions for this voxel. Grouping consecutively adjacent supported directions into the same continuous support region, and calculating... This represents the maximum solid angle covered on a unit sphere across all continuous support regions. Since the vehicle's motion relative to the base station in this embodiment is approximately within a horizontal plane, and all observation directions are approximately coplanar, for simplification, the observation directions can also be projected onto this horizontal plane, and the maximum width of the continuous support interval can be calculated based on the azimuth angle. ,at this time and The unit is degrees, not spherical degrees.

[0266] Step 7: Aggregation fraction for all voxels Sort and keep the first few. voxels; at the same time, it requires A voxel that satisfies both of these conditions is designated as occupied. The remaining voxels are set to unoccupied ( The final binary geometric reconstruction result is obtained. .

[0267] After the above steps, the occupied voxel The resulting structure is the geometric reconstruction of the vehicle. This result can be used for applications such as vehicle type recognition, vehicle classification, traffic flow monitoring, and vehicle type-based billing. This embodiment uses a roadside base station sensing a passing vehicle as an example, but this application is also applicable to other scenarios where stationary transceivers sense moving targets, such as indoor access points sensing pedestrians passing through corridors to reconstruct human body contours, and security systems sensing passing objects.

[0268] The third embodiment describes an application scenario where a mobile phone is moved indoors to reconstruct the geometry of the surrounding environment. See also... Figure 10 , Figure 10 This diagram illustrates a person walking with a mobile phone in an indoor corridor. The user holds a mobile phone equipped with a transceiver and walks normally indoors (e.g., in a room, corridor, shopping mall, etc.). During this movement, the phone continuously collects channel impulse responses at predetermined time intervals. Unlike the first embodiment where the transceiver observes an external target, in this embodiment, the transceiver is located inside the scene, and the region of interest (ROI) is the environment surrounding the transceiver (e.g., walls, furniture, pillars, etc.), meaning the ROI surrounds the transceiver's trajectory. The method described in this application reconstructs the three-dimensional geometry of the environment surrounding the mobile phone, which can be used for indoor map building, augmented reality navigation, indoor positioning assistance, etc. This embodiment corresponds to the sub-case of "ROI surrounding the transceiver trajectory" in motion configuration A of this application, with a single-station geometric configuration and a single transceiver system.

[0269] The following system parameters are used in this embodiment:

[0270] Transmitted signal: Native pilot signal or orthogonal frequency division multiplexing signal of mobile communication system, center frequency ,bandwidth Corresponding distance resolution ;

[0271] Observation configuration: A single handheld mobile phone was used to collect channel impulse responses at fixed time intervals (e.g., every few milliseconds or after moving a certain distance) while the user was walking indoors. A total of [number] impulse responses were collected. The location data is collected by the mobile phone's built-in inertial measurement unit, visual inertial odometry, or indoor positioning system.

[0272] Geometric configuration: Single-site, meaning the mobile phone's transmitting and receiving antennas are located in the same location. ;

[0273] Region of Interest: A rectangular area covering the environment surrounding the phone's movement trajectory, typically sized as follows. The phone's movement trajectory is located within this area;

[0274] Voxel mesh: 3D uniform mesh, voxel side length set to ,common Individual factors;

[0275] Parameters of the soft distance matching kernel function: Cut-off radius ,in Take typical scales of environmental characteristics, for example ,but ;

[0276] Threshold parameter: The aggregate score threshold is the aggregate score. Top voxels, angular support range threshold ;

[0277] Signal processing hardware: The processor built into the mobile phone (CPU / GPU / Neural Processing Unit (NPU)) can be used, or the collected data can be uploaded to the cloud for processing.

[0278] The specific implementation steps are as follows:

[0279] Step 1: The mobile phone transmits detection signals at 200 different locations during its movement indoors, and receives the echo signals reflected by the surrounding environment (walls, furniture, pillars, etc.). By performing correlation calculations between the received signals and known transmitted pilot signals, 200 sets of channel impulse responses are obtained. For each Square the amplitude to obtain the corresponding power delay spectrum. In this embodiment, the region of interest is the unknown environment surrounding the transceiver. There is no single known target distance, so the optional distance gating step (step 1A in the aforementioned embodiment) is not performed, and multipath components are directly extracted from the complete power delay spectrum.

[0280] Step 2: For each Peak detection is performed using a threshold-based or constant false alarm rate (CFAR) peak detection algorithm to extract local peaks in the power delay spectrum. Each peak corresponds to a multipath component generated by a scattering point in the surrounding environment, and its delay is extracted. and power .

[0281] Step 3: Select the region of interest Evenly divided into Voxel, recording each voxel center coordinates The various data collection locations of the mobile phone Located within the region of interest.

[0282] Step 4: For each voxel (voxel center location) ) and each observation Calculate the geometric distance based on the single-station configuration. Calculate the propagation distance of the response for each multipath component. .

[0283] Step 5: For each voxel, for each observation Each multipath component Calculate the distance difference metric Calculate the truncated Gaussian kernel. Calculate power weights Aggregation operations yield voxels. Aggregate fraction .

[0284] Step 6: Calculate the angular support range for each voxel. For each voxel, define its position on the first... The observation direction in this observation is a unit vector pointing from the voxel to the transceiver. When multipath components exist Make season Otherwise Examine the set of all supported observation directions for this voxel. Grouping consecutively adjacent supported directions into the same continuous support region, and calculating... This represents the maximum solid angle covered on a unit sphere across all continuously supported regions.

[0285] Step 7: Aggregation fraction for all voxels Sort and keep the first few. voxels; at the same time, it requires A voxel that satisfies both of these conditions is designated as occupied. The remaining voxels are set to unoccupied ( The final binary geometric reconstruction result is obtained. .

[0286] After the above steps, the occupied voxel The resulting structure is the geometric reconstruction of the environment surrounding the mobile phone. This result can be used for applications such as indoor map building, augmented reality navigation, indoor positioning assistance, and indoor robot navigation. This embodiment uses the reconstruction of the surrounding environment by the movement of a mobile phone indoors as an example, but this application is also applicable to other scenarios where transceivers move within a scene to reconstruct the surrounding environment, such as service robots or cleaning robots moving indoors to create maps, or drones flying inside tunnels or pipes to reconstruct the geometry of the inner walls.

[0287] Fourth embodiment: This embodiment describes an application scenario in which multiple dual-station transceivers in a room collaboratively sense a moving human body and reconstruct its geometric contours. See also... Figure 11, Figure 11 This diagram illustrates a system where multiple bi-station transceivers collaboratively sense a moving human body indoors. Multiple stationary transmitters and receivers are deployed at different locations within the room, with each transmitter and receiver forming multiple bi-station transceiver links. When a person moves within the room, each bi-station link continuously acquires channel impulse responses at multiple moments. Because the human body's position changes over time, and different bi-station links provide different geometric observation perspectives, a rich multi-view observation is achieved. The method described in this application reconstructs the three-dimensional geometric contour of the moving human body, which can be used for smart home human body sensing, indoor security, fall detection, and human behavior recognition. This embodiment corresponds to the motion configuration B (moving target + stationary transceiver) described in this application, with a bi-station geometric configuration and a multi-transceiver system.

[0288] The following system parameters are used in this embodiment:

[0289] Transmitted signal: Native pilot signal or orthogonal frequency division multiplexing signal at indoor access point, center frequency ,bandwidth Corresponding distance resolution ;

[0290] Observation configuration: Multiple transceivers, including 2 transmitters and 2 receivers, are stationary and deployed around the room at different locations. Each transmitter and each receiver form a two-station link. Two bi-station links; during human movement, each bi-station link collects channel impulse responses at fixed time intervals, collecting a total of [number missing] samples. At each moment, therefore the total number of observations ;

[0291] Geometric configuration: Dual-station, i.e., the transmitter locations for each link. With receiver position different;

[0292] Objective: A person moving around in a room, with a maximum horizontal dimension D of approximately... Typical walking speed is ;

[0293] Target location acquisition: The location of a person at each collection moment is reported to the transceiver by the mobile phone or other terminal on the person through its own positioning module.

[0294] Region of Interest: A cuboid region established with the human body as a reference, within a human body coordinate system, with typical dimensions... ;

[0295] Voxel mesh: 3D uniform mesh, voxel side length set to ,common Individual factors;

[0296] Parameters of the soft distance matching kernel function: Cut-off radius ;

[0297] Threshold parameter: The aggregate score threshold is the aggregate score. Top voxels, angular support range threshold ;

[0298] Signal processing hardware: can be achieved using indoor gateways, edge computing nodes' CPUs / GPUs, or by uploading data to the cloud for processing.

[0299] Specific implementation steps:

[0300] Step 1: During 30 data acquisition moments in the human movement process, four bistatic links transmit native pilot signals, and each receiver receives the echo signals reflected by the human body. By performing correlation calculations between the received signals and the known transmitted pilot signals, the desired results are obtained. Group channel impulse response For each Square the amplitude to obtain the corresponding power delay spectrum. .

[0301] Step 1A: Distance gating based on prior human distance. In this embodiment, the human body's position at each acquisition time is provided by a mobile phone or other terminal on the person through its own positioning module, thus distance gating can be performed. For each observation... Based on the known human body location and the transmitter location of the link and receiver position Determine the expected bistatic propagation distance window for the signal scattered by the human body. ,in and A spatial range slightly wider than the human body is obtained, converted into a time delay window, and only the power time delay spectrum is retained. Within this window, the portion outside the window is set to zero. After this gating, subsequent multipath component extraction is performed only on the time-delay segment belonging to the human body, excluding background scattering from furniture, walls, appliances, etc. in the room.

[0302] Step 2: For each Peak detection is performed using a threshold-based or constant false alarm rate (CFAR) peak detection algorithm to extract local peaks in the power delay spectrum. Each peak corresponds to a multipath component generated by a scattering point on the human body, and its delay is extracted. and power .

[0303] Step 3: Since the human body is in motion, to enable multi-view observations to be aggregated and calculated on a unified grid, a moving coordinate system is established with the human body as a reference. Using the human body's position at each acquisition time provided by the self-positioning module, the transmitter and receiver positions of each bistatic link at each acquisition time are transformed into the moving coordinate system, obtaining the... The equivalent transmitter position in the moving coordinate system is observed in the second observation. and equivalent receiver position The region of interest will be determined according to the human coordinate system. Evenly divided into Voxel, recording each voxel center coordinates .

[0304] Step 4: For each voxel (voxel center location) ) and each observation Calculate the geometric distance based on the dual-station configuration. Calculate the propagation distance of the response for each multipath component. .

[0305] Step 5: For each voxel, for each observation Each multipath component Calculate the distance difference metric Calculate the truncated Gaussian kernel. Calculate power weights Aggregation operations yield voxels. Aggregate fraction .

[0306] Step 6: Calculate the angular support range for each voxel. This embodiment uses a dual-station configuration, where each voxel is located at its [number]th station. Observation direction under this observation The bistatic bisector direction is adopted, i.e., from voxel to transmitter. The unit vector and the direction of the receiver The normalized direction of the sum of unit vectors. For each voxel, when multipath components exist. Make season Otherwise Examine all supported observational directions for this voxel. Grouping consecutively adjacent supported directions into the same continuous support region, and calculating... This represents the maximum solid angle covered on a unit sphere across all continuously supported regions.

[0307] Step 7: Aggregation fraction for all voxels Sort and keep the first few. voxels; at the same time, it requires A voxel that satisfies both of these conditions is designated as occupied. The remaining voxels are set to unoccupied ( The final binary geometric reconstruction result is obtained. .

[0308] After the above steps, the occupied voxel The resulting structure is the geometric reconstruction of a moving human body. This result can be used in applications such as smart home human body sensing, indoor security, fall detection, human behavior recognition, and non-contact health monitoring. This embodiment uses the example of multiple dual-station transceivers collaboratively sensing a moving human body in a room, but this application is also applicable to other scenarios involving multiple transceivers and dual-station configurations for sensing moving or stationary targets, such as multiple base stations collaboratively sensing vehicles and distributed access points collaboratively reconstructing the indoor environment.

[0309] The various methods of the embodiments of this application have been described above. Apparatus for implementing the above methods will now be provided.

[0310] Please refer to Figure 12 Another embodiment of this application provides a control device for environmental geometry reconstruction based on phaseless channel characteristics, comprising:

[0311] The first processing module 1201 is used to obtain phase-free channel features corresponding to multiple observations. The phase-free channel features are channel features that do not depend on phase information and can reflect the received signal strength at each propagation delay.

[0312] The second processing module 1202 is used to extract multipath components from the phaseless channel features of each observation to obtain the multipath components corresponding to each observation, wherein the multipath components include time delay and power.

[0313] The third processing module 1203 is used to obtain the distance difference metric between the geometric distance corresponding to each voxel in the voxel grid and the propagation distance corresponding to each multipath component in each observation, based on the geometric configuration information and the multipath components. The voxel grid is obtained by discretizing a preset region of interest.

[0314] The fourth processing module 1204 is used to perform aggregation operations on each voxel based on a soft distance matching kernel and power weight according to the multipath component and the distance difference metric, so as to obtain the aggregation score corresponding to each voxel.

[0315] The fifth processing module 1205 is used to obtain the angle support range corresponding to each voxel based on the geometric configuration information, the distance difference measurement and the preset cutoff radius.

[0316] The sixth processing module 1206 is used to perform voxel occupancy determination and geometric reconstruction based on the aggregation score and the angle support range, and obtain a binary geometric reconstruction result.

[0317] Specifically, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the phaseless channel characteristics include any one of the following: power delay spectrum, channel impulse response amplitude, and logarithmic power.

[0318] Specifically, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the extraction method for multipath components includes at least one of peak detection, constant false alarm rate detection, subspace method, sparse recovery, and deep learning.

[0319] Preferably, the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, before extracting the multipath component in a scenario where the phaseless channel characteristic is a power delay spectrum and the target position is known or estimable prior to the data, further includes:

[0320] The seventh processing module is used to obtain the expected propagation distance range of the signal scattered by the target;

[0321] The eighth processing module is used to perform spatiotemporal conversion based on the expected propagation distance range to obtain a time delay window;

[0322] The ninth processing module is used to filter the power delay spectrum according to the delay window to obtain the power delay spectrum for multipath component extraction.

[0323] Specifically, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the seventh processing module is specifically used for:

[0324] The expected propagation distance range is determined based on the target location and transceiver location, target tracking, motion model prediction, or pre-set methods.

[0325] Preferably, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the third processing module includes:

[0326] The first processing unit is configured to determine, based on the geometric configuration information, the sum of the distances from each voxel to the transmitter and receiver in each observation as the geometric distance;

[0327] The second processing unit is used to obtain the propagation distance corresponding to each multipath component based on the time delay corresponding to each multipath component in each observation;

[0328] The third processing unit is used to determine the difference between the geometric distance and the propagation distance in each observation as the distance difference metric;

[0329] Alternatively, the fourth processing unit is used to determine the ratio of the difference between the geometric distance and the propagation distance in each observation to the propagation distance as the distance difference metric.

[0330] Specifically, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the fourth processing module includes:

[0331] The fifth processing unit is used to obtain the soft distance matching kernel based on the distance difference metric and the preset cutoff radius in each observation. The soft distance matching kernel is a non-negative function that decreases monotonically as the absolute value of the distance difference metric increases, and is zero when the absolute value of the distance difference metric exceeds the preset cutoff radius.

[0332] The sixth processing unit is used to obtain the power weights corresponding to each multipath component according to a preset weight acquisition method.

[0333] The seventh processing unit is used to perform aggregation operations on the multipath components based on the soft distance matching kernel and the power weight to obtain the aggregation score corresponding to each voxel.

[0334] Preferably, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the soft range matching kernel is any one of the following: truncated Gaussian kernel, triangular kernel, double exponential kernel, inverse proportional kernel, and Epanechnikov kernel.

[0335] Preferably, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the preset weight acquisition method is any one of in-view power normalization, global power normalization, logarithmic weighting, square root weighting, and threshold weighting.

[0336] Specifically, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the aggregation operation includes: weighted summation, weighted averaging, summation in the logarithmic domain, or any other operation that can aggregate multiple said multipath components.

[0337] Preferably, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the fifth processing module includes:

[0338] The eighth processing unit is used to obtain the observation support status corresponding to each voxel based on the distance difference metric and the preset cutoff radius in each observation. If there is a multipath component whose absolute value of the distance difference metric is less than or equal to the preset cutoff radius, the observation support status is determined to be supported; otherwise, the observation support status is determined to be unsupported.

[0339] The ninth processing unit is used to determine the observation direction corresponding to each voxel in each observation based on the indication and location information of the single-station configuration or dual-station configuration in the geometric configuration information. If the observation support status corresponding to the voxel is supported in any observation, the observation direction of the voxel in that observation is stored in the observation direction set corresponding to the voxel.

[0340] The tenth processing unit is used to determine the angle support range corresponding to each voxel based on the set of observation directions.

[0341] Specifically, in the control device for environmental geometric reconstruction based on phaseless channel characteristics as described above, the angle support range is any one of the following: the maximum solid angle covered by the continuous support area on a unit sphere, the solid angle spanned by the convex hull of the observation direction, the total solid angle of the supported block, the proportion of the supported viewing angle, and the width of the continuous support interval after projection onto the main motion plane. The continuous support area is the area where the observation direction is continuously adjacent in the direction.

[0342] Preferably, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the sixth processing module includes:

[0343] The eleventh processing unit is used to obtain the aggregation score threshold and the angle support range threshold.

[0344] The twelfth processing unit is used to determine the score determination result based on the aggregated score and the aggregated score threshold.

[0345] The thirteenth processing unit is used to determine the angle determination result based on the angle support range and the angle support range threshold.

[0346] The fourteenth processing unit is used to determine the binary occupancy indicator corresponding to the target voxel based on the score determination result, the angle determination result, and the preset joint determination method. When the binary occupancy indicator corresponding to the target voxel is a first value, it indicates that the target voxel is occupied. When the binary occupancy indicator corresponding to the target voxel is a second value, it indicates that the target voxel is not occupied. The target voxel can be any one of the voxels.

[0347] The fifteenth processing unit is used to perform geometric reconstruction based on the binary occupancy indicators corresponding to all voxels, and obtain the binary geometric reconstruction result.

[0348] Specifically, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the eleventh processing unit is specifically used for:

[0349] Based on the aggregation scores, the data are sorted from largest to smallest, and the critical cumulative scores of the voxels at the first preset ratio are determined as the aggregation score threshold.

[0350] Alternatively, a pre-set absolute threshold can be obtained as the aggregation score threshold;

[0351] Alternatively, the aggregated score threshold can be adaptively determined based on a preset table relating the score threshold to motion configuration, geometric configuration, and / or quantity configuration.

[0352] Specifically, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the angle support range threshold is one of the following: a preset coverage solid angle value, a minimum number of consecutive supported viewing angles, and an angle coverage rate.

[0353] Specifically, the control device for environmental geometry reconstruction based on phaseless channel characteristics, as described above, uses one of the following joint determination methods: logical AND, weighted summation, product rule, or other determination methods that simultaneously reflect aggregate score and angle persistence.

[0354] Preferably, in the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above, the sixth processing module includes:

[0355] The sixteenth processing unit is used to input the aggregation score and the angle support range into the pre-trained voxel occupancy determination model, and obtain the binary occupancy indicator or occupancy probability corresponding to each voxel output by the voxel occupancy determination model. The voxel occupancy determination model is a model obtained by pre-training based on historical data and jointly determining the binary occupancy indicator or occupancy probability corresponding to the voxel based on the aggregation score and the angle support range.

[0356] The seventeenth processing unit is used to reconstruct the binary geometric reconstruction result based on the binary occupancy indication or occupancy probability corresponding to all the voxels.

[0357] Preferably, the control device for environmental geometry reconstruction based on phaseless channel characteristics as described above further includes:

[0358] The tenth processing module is used to render output based on the binary geometric reconstruction result, wherein the rendering output is at least one of geometric point cloud, three-dimensional point cloud, occupied grid, isosurface and triangular mesh.

[0359] The apparatus embodiments of this application are apparatuses corresponding to the embodiments of the methods described above. All implementation means in the method embodiments described above are applicable to the apparatus embodiments and can achieve the same technical effects. The apparatus provided in this application embodiments can implement all the method steps implemented in the method embodiments described above and can achieve the same technical effects. Therefore, the parts and beneficial effects that are the same as those in the method embodiments in this embodiment will not be described in detail here.

[0360] Another embodiment of this application provides an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described above, and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0361] Another embodiment of this application provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described above, and achieves the same technical effect. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0362] Another embodiment of this application provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described above, and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0363] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0364] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0365] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. An environmental geometry reconstruction method based on phase-free channel characteristics, characterized in that, include: The phase-free channel features corresponding to multiple observations are obtained. The phase-free channel features are channel features that do not depend on phase information and can reflect the received signal strength at each propagation delay. Multipath components are extracted from the phaseless channel features of each observation to obtain the multipath components corresponding to each observation. The multipath components include time delay and power. Based on the geometric configuration information and the multipath components, the distance difference metric between the geometric distance corresponding to each voxel in the voxel grid and the propagation distance corresponding to each multipath component is obtained in each observation. The voxel grid is obtained by discretizing a preset region of interest. Based on the multipath components and the distance difference metric, an aggregation operation based on a soft distance matching kernel and power weight is performed on each voxel to obtain the aggregation score corresponding to each voxel. Based on the geometric configuration information, the distance difference metric, and the preset cutoff radius, the angle support range corresponding to each voxel is obtained; Based on the aggregation score and the angle support range, voxel occupancy determination and geometric reconstruction are performed to obtain the binary geometric reconstruction result.

2. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 1, characterized in that, The phaseless channel characteristics include any one of the following: power delay spectrum, channel impulse response amplitude, and logarithmic power.

3. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 1, characterized in that, The extraction methods for the multipath components include at least one of peak detection, constant false alarm rate detection, subspace method, sparse recovery, and deep learning.

4. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 1, characterized in that, In the scenario where the phaseless channel feature is a power delay spectrum and the target position is known or estimable prior to the extraction of the multipath component, the method further includes: Obtain the expected propagation distance range of the signal after being scattered by the target; Based on the expected propagation distance range, a time delay window is obtained by performing a spatiotemporal conversion. The power delay spectrum is filtered according to the delay window to obtain the power delay spectrum used for multipath component extraction.

5. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 4, characterized in that, The expected propagation distance range of the acquired signal after being scattered by the target includes: The expected propagation distance range is determined based on the target location and transceiver location, target tracking, motion model prediction, or pre-set methods.

6. The environmental geometry reconstruction method based on phaseless channel characteristics according to claim 1, characterized in that, The step of obtaining the distance difference metric between the geometric distance corresponding to each voxel in the voxel grid and the propagation distance corresponding to each multipath component in each observation, based on the geometric configuration information and the multipath components, includes: Based on the geometric configuration information, the sum of the distances from each voxel to the transmitter and receiver in each observation is determined as the geometric distance; Based on the time delay corresponding to each of the multipath components in each observation, the propagation distance corresponding to each of the multipath components is obtained; The difference between the geometric distance and the propagation distance in each observation is defined as the distance difference metric. Alternatively, the distance difference metric can be determined as the ratio of the difference between the geometric distance and the propagation distance in each observation to the propagation distance.

7. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 1, characterized in that, The step of performing aggregation operations on each voxel based on a soft distance matching kernel and power weights according to the multipath components and the distance difference metric to obtain the aggregation score corresponding to each voxel includes: The soft distance matching kernel is obtained based on the distance difference metric and the preset cutoff radius in each observation. The soft distance matching kernel is a non-negative function that decreases monotonically as the absolute value of the distance difference increases, and is zero when the absolute value of the distance difference exceeds the preset cutoff radius. According to the preset weight acquisition method, the power weight corresponding to each multipath component is obtained; Based on the soft distance matching kernel and the power weight, the multipath components are aggregated to obtain the aggregated score corresponding to each voxel.

8. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 1 or 7, characterized in that, The soft distance matching kernel is any one of the following: truncated Gaussian kernel, triangular kernel, double exponential kernel, inverse proportional kernel, and Epanechnikov kernel.

9. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 7, characterized in that, The preset weight acquisition method is any one of the following: in-view power normalization, global power normalization, logarithmic weighting, square root weighting, and threshold weighting.

10. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 7, characterized in that, The aggregation operations include: weighted summation, weighted average, summation in the logarithmic field, or any other operation that can aggregate multiple of the multipath components.

11. The environmental geometry reconstruction method based on phaseless channel characteristics according to claim 1, characterized in that, The step of obtaining the angle support range corresponding to each voxel based on the geometric configuration information, the distance difference metric, and the preset cutoff radius includes: Based on the distance difference metric and the preset cutoff radius in each observation, the observation support status corresponding to each voxel is obtained. If there is a multipath component whose absolute value of the distance difference metric is less than or equal to the preset cutoff radius, the observation support status is determined to be supported; otherwise, the observation support status is determined to be unsupported. Based on the single-station configuration or dual-station configuration indication and location information in the geometric configuration information, the observation direction corresponding to each voxel in each observation is determined. If the observation support status corresponding to the voxel is supported in any observation, the observation direction of the voxel in that observation is stored in the observation direction set corresponding to the voxel. Based on the set of observation directions, the angular support range corresponding to each voxel is determined.

12. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 11, characterized in that, The angle support range is any one of the following: the maximum solid angle covered by the continuous support area on the unit sphere determined based on the set of observation directions, the solid angle spanned by the convex hull of the observation direction, the total solid angle of the supported block, the proportion of the supported viewing angle, and the width of the continuous support interval after projection onto the main motion plane. The continuous support area is the area where the observation directions are consecutively adjacent in the direction.

13. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 1, characterized in that, The process of determining voxel occupancy and performing geometric reconstruction based on the aggregation score and the angular support range to obtain a binary geometric reconstruction result includes: Obtain the aggregate score threshold and the angle support range threshold; The score determination result is determined based on the aggregated score and the aggregated score threshold. The angle determination result is determined based on the angle support range and the angle support range threshold. Based on the score determination result, the angle determination result, and the preset joint determination method, the binary occupancy indicator corresponding to the target voxel is determined. When the binary occupancy indicator corresponding to the target voxel is a first value, it indicates that the target voxel is occupied. When the binary occupancy indicator corresponding to the target voxel is a second value, it indicates that the target voxel is not occupied. The target voxel can be any one of the voxels. Geometric reconstruction is performed based on the binary occupancy indicators corresponding to all voxels to obtain the binary geometric reconstruction result.

14. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 13, characterized in that, The process of obtaining the aggregate score threshold includes: Based on the aggregation scores, the data are sorted from largest to smallest, and the critical cumulative scores of the voxels at the first preset ratio are determined as the aggregation score threshold. Alternatively, a pre-set absolute threshold can be obtained as the aggregation score threshold; Alternatively, the aggregated score threshold can be adaptively determined based on a preset table relating the score threshold to motion configuration, geometric configuration, and / or quantity configuration.

15. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 13, characterized in that, The angle support range threshold is one of the following: a preset solid angle value, a minimum number of consecutive supported viewing angles, and an angle coverage rate.

16. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 13, characterized in that, The joint determination method is one of the following: logical AND, weighted summation, product rule, or other determination methods that simultaneously reflect aggregate score and angle persistence.

17. The environmental geometry reconstruction method based on phase-free channel characteristics according to claim 1, characterized in that, The process of determining voxel occupancy and performing geometric reconstruction based on the aggregation score and the angular support range to obtain a binary geometric reconstruction result includes: The aggregation score and the angle support range are input into the pre-trained voxel occupancy determination model to obtain the binary occupancy indicator or occupancy probability corresponding to each voxel output by the voxel occupancy determination model. The voxel occupancy determination model is a model that is pre-trained based on historical data and jointly determines the binary occupancy indicator or occupancy probability corresponding to the voxel based on the aggregation score and the angle support range. The binary geometric reconstruction result is obtained by reconstructing based on the binary occupancy indicator or occupancy probability corresponding to all the voxels.

18. The environmental geometry reconstruction method based on phaseless channel characteristics according to claim 1, characterized in that, Also includes: The rendering output is based on the binary geometric reconstruction result, wherein the rendering output is at least one of geometric point cloud, 3D point cloud, occupied grid, isosurface and triangular mesh.

19. A control device for environmental geometry reconstruction based on phaseless channel characteristics, characterized in that, include: The first processing module is used to obtain phase-free channel features corresponding to multiple observations. The phase-free channel features are channel features that do not depend on phase information and can reflect the received signal strength at each propagation delay. The second processing module is used to extract multipath components from the phaseless channel features of each observation to obtain the multipath components corresponding to each observation. The multipath components include time delay and power. The third processing module is used to obtain the distance difference metric between the geometric distance of each voxel in the voxel grid and the propagation distance of each multipath component in each observation, based on the geometric configuration information and the multipath components. The voxel grid is obtained by discretizing a preset region of interest. The fourth processing module is used to perform aggregation operations on each voxel based on a soft distance matching kernel and power weight according to the multipath components and the distance difference metric, so as to obtain the aggregation score corresponding to each voxel; The fifth processing module is used to obtain the angle support range corresponding to each voxel based on the geometric configuration information, the distance difference metric, and the preset cutoff radius. The sixth processing module is used to perform voxel occupancy determination and geometric reconstruction based on the aggregation score and the angle support range, and obtain the binary geometric reconstruction result.

20. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described in any one of claims 1 to 18.

21. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described in any one of claims 1 to 18.

22. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the environmental geometry reconstruction method based on phaseless channel characteristics as described in any one of claims 1 to 18.

Images