Multi-sensor data fusion method and system based on intelligent pod

Abstract

The present application provides a multi-sensor data fusion method and system based on an intelligent pod. The method collects the radiation data stream of the multi-spectral sensor of the intelligent pod under the electromagnetic interference field. Based on the data, the radiation transfer equation is used to calculate the quantitative attenuation coefficient of the radiation value of each band in the interference field. The coefficient is used to generate the directional beam of the phased array antenna, and the target area is scanned to obtain echo data. Finally, the original radiation data stream and the target echo data are jointly sparsely represented to generate a fusion result. Multi-sensor data fusion based on intelligent pods. The present application quantifies the attenuation of electromagnetic interference on radiation in each band, dynamically generates an anti-interference beam to scan the target, and combines the joint sparse representation to fuse multi-source data to improve the perception performance of the intelligent pod in a complex electromagnetic environment.

Description

Technical Field

[0001] The present application relates to the technical field of multi-sensor data fusion, and in particular to a multi-sensor data fusion method and system based on an intelligent pod. Background Art

[0002] When intelligent pods perform reconnaissance and surveillance missions in complex, high-electromagnetic interference environments, the data acquisition and fusion of their onboard multispectral and radar sensors face significant challenges. Electromagnetic interference can severely attenuate multispectral radiation signals and contaminate radar echoes, leading to distortion or failure of single sensor data. Therefore, a method is urgently needed to effectively fuse multispectral and radar data in strong electromagnetic interference fields. The core requirement is to quantify the attenuation effect of electromagnetic interference on different optical bands and dynamically optimize radar perception strategies accordingly, ultimately achieving highly robust multi-source information fusion and improving the accuracy and reliability of target detection and recognition.

[0003] A representative solution for this type of scenario is radar-optical time-sharing fusion based on adaptive beamforming. This approach first uses the intelligent pod's optical sensors to obtain rough scene information during periods of interference or through preliminary filtering, identifying potential areas of interest. Subsequently, the phased array radar is controlled to adaptively adjust the direction and parameters of the transmit beam based on this preliminary optical information, performing an anti-interference optimized scan of the area of ​​interest to acquire target echo data. Finally, at the feature layer or decision layer, the optical recognition results are fused with radar detection data, such as range, velocity, and high-resolution imaging, to output final target information.

[0004] However, while these existing solutions introduce adaptive beamforming to improve radar anti-interference capabilities, they still suffer from significant drawbacks: First, optical perception and radar scanning are performed in phases, preventing the simultaneous utilization of multi-source information. In dynamic, high-interference environments where interference patterns rapidly change, environmental changes between optical information acquisition and radar scanning can render optimized beam parameters ineffective, resulting in the loss of optimal detection opportunities. Second, these solutions fail to deeply quantify the dynamic and precise attenuation effects of electromagnetic interference on radiation data across multiple spectral bands. They only utilize preliminary optical information to guide the radar, failing to incorporate the quantified attenuation coefficient of electromagnetic interference on the optical channel as a key input to guide radar beam formation and optimization, limiting the upper limit of anti-interference and perception accuracy. Third, they primarily perform post-fusion at the feature or decision layer, failing to perform deep joint processing at the lower-level raw data or sparse representation layers. This limits the ability to leverage the potential complementarity and correlation between multi-source data to jointly suppress interference and enhance the effective signal. This is particularly true when the raw data are all contaminated to varying degrees by interference, making the fusion effect susceptible to noise. Summary of the Invention

[0005] The present application provides a multi-sensor data fusion method and system based on an intelligent pod to solve the problem of poor multi-sensor data fusion effect in the prior art.

[0006] In a first aspect, the present application provides a multi-sensor data fusion method based on an intelligent pod, comprising:

[0007] Collecting radiation data streams generated by the multispectral sensor of the smart pod in an electromagnetic interference field, wherein the radiation data streams include radiation data in the visible light band, near infrared band, and thermal infrared band;

[0008] Based on the radiation data stream, calculating the quantitative attenuation coefficient of the radiation value of each band in the electromagnetic interference field by using a radiation transfer equation;

[0009] Based on the quantized attenuation coefficient, a directional beam corresponding to the phased array antenna array in the intelligent pod is generated, and the directional beam is used to scan the target area to obtain target echo data;

[0010] The radiation data stream and the target echo data are jointly sparsely represented to generate a multi-sensor data fusion result.

[0011] Optionally, generating a directional beam corresponding to the phased array antenna array in the smart pod based on the quantized attenuation coefficient, and scanning a target area using the directional beam to obtain target echo data includes:

[0012] Determining adjustment parameters of each antenna element of a phased array antenna array in the smart pod based on the quantized attenuation coefficient, wherein the adjustment parameters include a phase offset and a signal amplitude weight;

[0013] According to the adjustment parameters, corresponding phase offsets and signal amplitude weights are simultaneously applied to all antenna elements of the phased array antenna array to generate a directional beam corresponding to the phased array antenna array in the smart pod;

[0014] The combination of phase offset and signal amplitude weight is repeatedly changed so that the directional beam is directed to different sub-areas of the target area in turn. In the direction of each sub-area, a detection signal is transmitted through the phased array antenna array, and a reflected signal from the target area is received as target echo data.

[0015] Optionally, performing a joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result includes:

[0016] Combining the radiation data stream and the target echo data into a multi-dimensional data set; and constructing a common representation space using a set of basis vectors;

[0017] In the common representation space, solving a sparse coefficient vector of the multidimensional data set, wherein the sparse coefficient vector satisfies a minimization of data reconstruction error and a sparsity constraint;

[0018] Based on the sparse coefficient vector, a fused data sequence representation is reconstructed as a multi-sensor data fusion result.

[0019] Optionally, based on the quantized attenuation coefficient, determining adjustment parameters of each antenna element of the phased array antenna array in the smart pod, wherein the adjustment parameters include a phase offset and a signal amplitude weight, comprises:

[0020] Obtaining the physical position of each antenna unit in the phased array antenna array as a position coordinate;

[0021] An initial phase value is obtained by performing a dot product operation on the position coordinates of the antenna unit and a reference direction vector, and then the initial phase value is multiplied by a constant factor related to the signal wavelength to obtain a basic phase offset;

[0022] Inputting the basic phase offset into a predefined phase adjustment function, using the phase adjustment function to take the quantized attenuation coefficient as an input variable, outputting a phase correction factor, and multiplying the basic phase offset by the phase correction factor to obtain a phase offset;

[0023] Assigning a weight related to the distance from the array center to each antenna element based on the relative position of the antenna element in the phased array antenna array using a predefined position weight function to obtain a basic signal amplitude weight;

[0024] The basic signal amplitude weight is input into a predefined amplitude adjustment function, and the amplitude adjustment function outputs an amplitude scaling factor with the quantized attenuation coefficient as an input variable, and the basic signal amplitude weight is multiplied by the amplitude scaling factor to obtain the signal amplitude weight.

[0025] Optionally, solving the sparse coefficient vector of the multidimensional data set in the common representation space includes:

[0026] constructing an optimization objective based on the multidimensional data set and a set of basis vectors in the common representation space, the optimization objective comprising a reconstruction error term and a sparsity constraint term, wherein the reconstruction error term represents a difference between the multidimensional data set and a linear combination of the basis vectors, and the sparsity constraint term represents a control of the number of non-zero elements of the sparse coefficient vector; solving the optimization objective through a preset iterative process to determine a sparse coefficient vector of the multidimensional data set;

[0027] The preset iterative process includes: initializing the sparse coefficient vector as a starting vector; in each iteration, updating each element of the sparse coefficient vector in turn, wherein the update of each element is based on the joint influence of the current reconstruction error and the sparsity constraint term, and checking whether the updated sparse coefficient vector meets the preset convergence condition, if so, terminating the iteration, otherwise continuing the iteration;

[0028] When each element of the sparse coefficient vector is updated, an intermediate value is calculated based on the basis vector and the multidimensional data set, and the intermediate value is compared with a regularization parameter to determine the final value of the element; the regularization parameter is used to balance the weights of the reconstruction error term and the sparsity constraint term, and remains unchanged during the iterative optimization process; the preset convergence condition is based on whether the change in the sparse coefficient vector in adjacent iterations is lower than a fixed threshold.

[0029] Optionally, reconstructing a fused data sequence representation based on the sparse coefficient vector as a multi-sensor data fusion result includes:

[0030] According to the sparse coefficient vector and a predefined set of basis vectors, multiplying each basis vector by a corresponding coefficient element in the sparse coefficient vector to generate a set of weighted basis vectors;

[0031] Superimposing all the weighted basis vectors one by one according to element positions to form a fused data sequence;

[0032] The fused data sequence is divided according to the characteristic representation lengths of the radiation data stream and the target echo data, and a multi-sensor data fusion result is output.

[0033] Optionally, based on the radiation data stream, calculating the quantized attenuation coefficient of the radiation value of each band in the electromagnetic interference field by using a radiation transfer equation includes:

[0034] For each spectral band in the radiation data stream, extracting a reference radiation value of the spectral band in a field without electromagnetic interference;

[0035] Inputting the reference radiation value and the actual radiation value under the electromagnetic interference field into the radiation transfer equation to calculate the radiation attenuation ratio of each band;

[0036] The radiation attenuation ratio is numerically quantified and the output is a quantized attenuation coefficient.

[0037] In a second aspect, the present application provides a multi-sensor data fusion system based on an intelligent pod, comprising:

[0038] The first generation module collects the radiation data stream generated by the multispectral sensor of the intelligent pod in the electromagnetic interference field, and the radiation data stream includes radiation data in the visible light band, the near infrared band, and the thermal infrared band;

[0039] A calculation module, which calculates the quantitative attenuation coefficient of the radiation value of each band in the electromagnetic interference field through a radiation transfer equation based on the radiation data stream;

[0040] A second generating module generates a directional beam corresponding to the phased array antenna array in the intelligent pod based on the quantized attenuation coefficient, and uses the directional beam to scan the target area to obtain target echo data;

[0041] The third generation module performs a joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result.

[0042] In an embodiment of the present application, a radiation data stream generated by a multispectral sensor of an intelligent pod in an electromagnetic interference field is collected, where the radiation data stream includes radiation data in a visible light band, a near-infrared band, and a thermal infrared band. Based on the radiation data stream, a quantized attenuation coefficient of the radiation value of each band in the electromagnetic interference field is calculated using a radiation transfer equation. Based on the quantized attenuation coefficient, a directional beam corresponding to the phased array antenna array in the intelligent pod is generated, and the directional beam is used to scan the target area to obtain target echo data. The radiation data stream and the target echo data are jointly sparsely represented to generate a multi-sensor data fusion result.

[0043] The technical solution of this application has the following beneficial effects:

[0044] Acquire multi-band optical radiation data in electromagnetic interference environments, providing raw input for subsequent quantification of interference impacts, ensuring data coverage across diverse spectral characteristics and improving comprehensive environmental perception. Accurately quantify the degree of attenuation of electromagnetic interference on radiation in different bands, providing key parameters for subsequent phased array beam optimization and enhancing anti-interference capabilities. Dynamically adjust the phased array antenna's beam parameters (such as phase offset and amplitude weighting) using attenuation coefficients to enhance radar signal directivity and signal-to-noise ratio in interference environments, improving target detection accuracy. Optimize sparse fusion of multi-source data within a common representation space to reduce redundant information, enhance data complementarity, and improve target recognition and anti-interference capabilities.

[0045] Furthermore, based on the quantized attenuation coefficient, the phase offset and signal amplitude weight of the phased array antenna unit are dynamically adjusted to generate a directional beam. By changing the parameter combination, the beam is made to scan different sub-areas of the target area to obtain high-precision echo data. Subsequently, the multispectral radiation data and the radar echo data are combined into a multidimensional data set, and sparse optimization is performed in a common representation space to solve the sparse coefficient vector that minimizes the reconstruction error, and finally a fused data sequence is generated.

[0046] This method improves the radar's directivity and anti-interference capabilities in strong interference environments by dynamically optimizing phased array beam parameters. At the same time, it uses joint sparse representation technology to achieve deep fusion of multi-source information at the underlying data level, reduce the impact of noise, and enhance the robustness and recognition accuracy of target features. It is suitable for collaborative perception of multiple sensors in intelligent pods in complex electromagnetic environments.

[0047] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0049] Figure 1 A flowchart of a multi-sensor data fusion method based on an intelligent pod provided by the present application is shown;

[0050] Figure 2 The present invention provides a schematic structural diagram of a multi-sensor data fusion system based on an intelligent pod;

[0051] Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0052] In order to enable people skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0053] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0054] Existing multi-sensor fusion technology for intelligent pods faces significant limitations in strong electromagnetic interference environments. While adaptive radar time-sharing fusion schemes based on optical guidance can initially optimize beam pointing, their time-sharing operation mode leads to spatiotemporal mismatches between optical and radar data, making them incapable of responding to dynamic interference changes. More critically, existing methods fail to quantify the differential attenuation effects of electromagnetic interference on various multispectral bands, relying solely on coarse radar parameter adjustments based on distorted optical data. This lacks a precise physical basis for beam anti-interference optimization. Furthermore, traditional feature-layer / decision-layer fusion approaches struggle to eliminate interference noise from multi-source data at the underlying level, resulting in insufficient robustness in the fusion results. These shortcomings stem from a lack of understanding of electromagnetic interference mechanisms and the rigidity of multi-sensor data coordination mechanisms.

[0055] To address these issues, this paper proposes a multi-sensor fusion method for intelligent pods based on quantized attenuation modeling and joint sparse representation. Its core innovation lies in: first, using the radiative transfer equation to accurately calculate the quantized attenuation coefficients for the visible, near-infrared, and thermal infrared bands in the electromagnetic interference field, establishing an interference-band correlation model. This coefficient is then used as a physical constraint for phased array beamforming, dynamically adjusting the phase offsets and amplitude weights of antenna elements to achieve a closed-loop optimization of the directional beam scanning parameters and optical attenuation characteristics. Finally, through joint sparse reconstruction of multi-source data in a common representation space, deep coupling of optical and radar data at the raw signal level is achieved. This method overcomes the inherent limitations of time-sharing operation. The quantized attenuation coefficients provide a unified interference characterization benchmark for optical and radar systems, enabling them to respond collaboratively to environmental changes. Furthermore, the joint sparse representation effectively suppresses interference noise in single-sensor data by enhancing the complementarity of the underlying data, addressing the limited anti-interference capability inherent in traditional fusion methods due to their shallow layers. Experiments demonstrate that this approach maintains high target recognition accuracy even in strong interference fields, significantly improving upon existing technologies.

[0056] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0057] Figure 1 A flowchart of a multi-sensor data fusion method based on an intelligent pod is provided for an embodiment of the present application. Figure 1 As shown, the method includes:

[0058] 101. Collecting a radiation data stream generated by a multispectral sensor of the intelligent pod in an electromagnetic interference field, wherein the radiation data stream includes radiation data in a visible light band, a near infrared band, and a thermal infrared band;

[0059] In the above solution, the smart pod is an airborne device equipped with multiple sensors, using a three-axis stabilized platform to ensure data acquisition stability. A multispectral sensor simultaneously captures electromagnetic radiation from a target area, with its output divided into three independent channels: visible light (380–750nm, reflecting object color and texture), near-infrared (750–1300nm, used for vegetation analysis), and thermal infrared (8–14μm, detecting surface temperature). The radiation data stream is a continuous time series data set formed by aligning the visible light, near-infrared, and thermal infrared data by timestamp, ensuring spatiotemporal synchronization of data across different bands.

[0060] In an embodiment of the present application, step 101 synchronously captures radiation information from the target area through the multispectral sensor of the intelligent pod in an electromagnetic interference field environment; the radiation information is separated into three independent data channels according to the spectral band: the visible light band, the near-infrared band, and the thermal infrared band; the data streams of the three independent data channels are integrated into a continuous time-series data stream to form a radiation data stream.

[0061] For example, in a farmland monitoring mission at site A, an intelligent pod equipped with a multispectral sensor flew to the airspace near the high-voltage substation B (a strong electromagnetic interference field).

[0062] Specifically, the pod synchronously collects radiation information from the farmland below at a 10Hz frequency. The visible light channel captures crop color information, the near-infrared channel measures vegetation moisture content, and the thermal infrared channel captures soil temperature distribution. After time alignment, the three-channel data is integrated into a continuous data stream containing timestamps, band identifiers, and radiation intensity, and transmitted to a ground processing station.

[0063] This step achieves high-precision synchronous acquisition and spatiotemporal alignment of multispectral data in an electromagnetic interference environment, providing standardized input for subsequent attenuation analysis and avoiding analysis deviations caused by band separation or timing misalignment.

[0064] 102. Calculate the quantitative attenuation coefficient of the radiation value of each band in the electromagnetic interference field using a radiation transfer equation based on the radiation data stream;

[0065] Optionally, according to step 102, based on the radiation data stream, calculating the quantized attenuation coefficient of the radiation value of each band in the electromagnetic interference field by using a radiation transfer equation includes:

[0066] 1021. For each spectral band in the radiation data stream, extract a reference radiation value of the spectral band in a field without electromagnetic interference;

[0067] 1022. Input the reference radiation value and the actual radiation value under the electromagnetic interference field into the radiation transfer equation to calculate the radiation attenuation ratio of each band;

[0068] 1023. Numerically quantify the radiation attenuation ratio, and output the quantized attenuation coefficient.

[0069] In the above solution, the radiation data stream refers to electromagnetic wave energy data collected by the intelligent pod's multispectral sensor. It covers three wavelength bands: visible light (400-700nm), near-infrared (700-1300nm), and thermal infrared (8-14μm), representing the radiation intensity of the target scene. The reference radiation value refers to the pre-calibrated baseline radiation intensity for each wavelength band under an ideal environment free of electromagnetic interference, serving as a reference for attenuation calculations. The actual radiation value refers to the radiation intensity for each wavelength band collected in an electromagnetic interference field, including signal distortion caused by interference. The radiation attenuation ratio refers to the degree of interference attenuation calculated using the radiation transfer equation. The quantitative attenuation coefficient is a normalized value (ranging from 0 to 1) of the attenuation ratio, used to quantify the differential impact of interference on each wavelength band.

[0070] In this embodiment, step 1021 first calls a pre-stored reference radiation value database to match the radiation data stream from step 101 by spectral band. Using timestamp alignment technology, calibration data from the same geographic location and lighting conditions in an electromagnetic interference-free environment are associated with the data stream to extract reference radiation values ​​for the visible, near-infrared, and thermal infrared bands.

[0071] Secondly, in step 1022, the reference radiation value extracted in step 1011 and the actual radiation value under the electromagnetic interference field are input into the radiation transfer equation to establish a mathematical model: ,in is the reference radiation value without interference field, which comes from laboratory calibration data; It is the measured radiation value of the electromagnetic interference field, which comes from the radiation data stream; is the band-related attenuation coefficient, which characterizes the attenuation intensity of electromagnetic waves in the interference field. is the radiation propagation path length, obtained through the pod positioning system; is the center wavelength of the current processing band. Calculate the band-related attenuation coefficient according to the model and bring it into the calculation formula: , we can conclude The value of is the radiation attenuation ratio of each band.

[0072] Finally, step 1023 is used to clean and smooth the radiation attenuation ratio, and instantaneous noise is eliminated by automatically returning negative values ​​to zero, cutting off excess values, and performing weighted moving average filtering. Subsequently, cross-band time calibration is performed to align the time series of near-infrared and thermal infrared with visible light as the reference. The processed attenuation ratio is then linearly mapped to a 0-255 integer value space. Finally, it is encapsulated into a standardized data packet containing a timestamp, band identifier, and quantized value, and the output is a quantized attenuation coefficient that can be directly used for multi-sensor fusion.

[0073] For example, when monitoring crop growth in industrial area A, the smart pod flew over the high-voltage power grid of Plant B (a source of electromagnetic interference).

[0074] Specifically, reference radiation values ​​for the undisturbed farmland area C are first retrieved, for example, visible light: 1500 Lux, near infrared: 2.3 W / m², and thermal infrared: 310 K. During flight, interference data is collected, for example, visible light: 920 Lux, near infrared: 1.2 W / m², and thermal infrared: 298 K. The results are calculated using the radiation transfer equation and normalized to derive the quantified attenuation coefficient. This process yields the quantified attenuation coefficients for the visible light band, near infrared band, and thermal infrared band. The calculation process is as follows: Visible light band: , , , the visible light attenuation ratio is 0.434, and the normalized quantitative attenuation coefficient is 0.42; in the near infrared band: , , the attenuation ratio of the near-infrared band is 0.510, and the normalized quantitative attenuation coefficient is 0.51; thermal infrared band: , , , the attenuation ratio of the thermal infrared band is 0.142, and the quantized attenuation coefficient is 0.14 after normalization. Finally, the calculation results are output in sequence to form the basis for anti-interference compensation.

[0075] This step achieves the precise quantification of multi-spectral radiation attenuation in the electromagnetic interference field. By establishing a physical model association between the reference benchmark and the data, the complex electromagnetic environment impact is converted into a calculable band-independent coefficient, providing an anti-interference correction basis for subsequent multi-sensor data fusion, significantly improving the reliability of the fused data and the target recognition accuracy in a strong interference environment.

[0076] 103. Based on the quantized attenuation coefficient, generate a directional beam corresponding to the phased array antenna array in the smart pod, and use the directional beam to scan the target area to obtain target echo data;

[0077] Optionally, in step 103, generating a directional beam corresponding to the phased array antenna array in the smart pod based on the quantized attenuation coefficient, and using the directional beam to scan the target area to obtain target echo data includes:

[0078] 1031. Determine adjustment parameters of each antenna element of the phased array antenna array in the smart pod based on the quantized attenuation coefficient, wherein the adjustment parameters include a phase offset and a signal amplitude weight;

[0079] Among them, the process of step 1031 "determining the adjustment parameters of each antenna unit of the phased array antenna array in the smart pod based on the quantized attenuation coefficient, wherein the adjustment parameters include a phase offset and a signal amplitude weight" includes: obtaining the physical position of each antenna unit in the phased array antenna array as a position coordinate; performing a dot product operation on the position coordinate of the antenna unit and a reference direction vector to obtain an initial phase value, and then multiplying the initial phase value by a constant factor related to the signal wavelength to obtain a basic phase offset; inputting the basic phase offset into a predefined phase adjustment function, using the quantized attenuation coefficient as an input variable through the phase adjustment function to output a phase correction factor, and multiplying the basic phase offset by the phase correction factor to obtain a phase offset; assigning a weight related to the array center distance to each antenna unit based on the relative position of the antenna unit in the phased array antenna array through a predefined position weight function to obtain a basic signal amplitude weight; inputting the basic signal amplitude weight into a predefined amplitude adjustment function, using the quantized attenuation coefficient as an input variable through the amplitude adjustment function to output an amplitude scaling factor, and multiplying the basic signal amplitude weight by the amplitude scaling factor to obtain a signal amplitude weight.

[0080] 1032. Apply corresponding phase offsets and signal amplitude weights to all antenna elements of the phased array antenna array according to the adjustment parameters to generate a directional beam corresponding to the phased array antenna array in the smart pod;

[0081] 1033. Repeatedly change the combination of phase offset and signal amplitude weight so that the directional beam is directed to different sub-areas of the target area in turn, and transmit a detection signal through the phased array antenna array in the direction of each sub-area, and receive a reflected signal from the target area as target echo data.

[0082] In the above scheme, the quantized attenuation coefficient is a normalized value (ranging from 0 to 1) that quantifies the differential impact of interference on different bands. A directional beam is a directional electromagnetic beam formed by controlling the phase and amplitude of each element of a phased array antenna. Target echo data refers to the strength and delay information of the reflected signal received after the directional beam scans the target. The phase offset is the adjustment parameter (in radians) that controls the phase angle of the transmitted signal from the antenna element. The signal amplitude weight is the gain factor that adjusts the transmit power of the antenna element. The position coordinates are the physical location (x, y, z) of the antenna element in three-dimensional space. The reference direction vector is the unit direction vector (θ, φ) of the desired beam direction. The phase correction factor is a phase compensation value calculated from the quantized attenuation coefficient to offset the effects of environmental attenuation. The amplitude scaling factor is an amplitude compensation value calculated from the quantized attenuation coefficient to increase gain in weak signal areas.

[0083] In this embodiment of the present application, step 1031 first calls the pre-stored structure database of the phased array antenna array to extract the three-dimensional position coordinates (x, y, z) of each antenna element in the pod coordinate system, and establishes a position matrix containing the spatial distribution of all elements. For example, element A03 in an 8×8 array corresponds to coordinates (0.12m, -0.18m, 0.05m). A reference direction vector is then calculated based on the preset azimuth / elevation angles of the current target sub-area being scanned, for example, 310° in azimuth and 20° in elevation. A spatial projection operation is then performed on the position coordinates of each antenna element and this vector to obtain an initial phase value representing the wavefront delay. This initial phase value is then multiplied by a fixed scaling factor determined by the operating frequency, for example, 25.12 rad / m at a frequency of 2.4 GHz, to output the uncorrected base phase offset. For example, an array edge element has a base offset of -1.3 rad due to propagation path differences. The current quantized attenuation coefficient, for example, the three-band value, is then input into a preset phase adjustment function. This function performs a hierarchical process: first, the three-band coefficients are weighted averaged, i.e., the thermal infrared weight is 0.6, the near infrared weight is 0.3, and the visible light weight is 0.1, to obtain the comprehensive attenuation value; then, the phase correction factor is calculated by piecewise linear interpolation, and the output is 1.0 when the comprehensive attenuation value is <0.3, and 1 + (comprehensive attenuation value -0.3) when it is between 0.3 and 0.7. The factor increases by 0.5 and remains constant at 1.2 when it is greater than 0.7. The base phase offset is multiplied by this factor to generate the actual phase offset. Secondly, based on the Euclidean distance from the antenna element to the center of the array, a Gaussian distribution weight is assigned: the center element has a weight of 1.0, and the weight decreases by 10% for every 0.1m increase in distance. For example, a element 0.25m from the center receives a base weight of 0.78. Finally, the thermal infrared attenuation dominance is mapped to a scaling factor. When the thermal infrared attenuation dominance is <0.4, the factor = 1.0, and when it is 0.4~0.8, the factor = 1.0-(thermal infrared attenuation dominance-0.4) 0.75, >0.8 when the factor = 0.7. The base amplitude weight is multiplied by this factor to get the final signal amplitude weight, for example, base weight 0.78 × factor 0.85 = 0.663.

[0084] Secondly, through step 1032, the phase offset of each antenna unit, that is, the waveform delay parameter and signal amplitude weight dynamically corrected based on the target direction and the quantized attenuation coefficient, that is, the power scaling coefficient calculated according to the array position and the attenuation strength, is synchronously distributed to all units of the phased array through a high-speed bus; the digital phase shifter of each unit rotates the starting waveform point of the carrier signal according to the phase offset, and the digitally controlled attenuator adjusts the transmission power according to the signal amplitude weight, so that all units generate electromagnetic waves with consistent frequency but differentiated phase / amplitude under a unified clock signal; these beams produce an interference effect in space, that is, a coherently enhanced main lobe is formed in the preset target direction due to the alignment of the wavefront, and other directions are suppressed due to the random superposition and cancellation of the wavefronts, and finally a sharp directional beam is synthesized with a pointing accuracy better than 0.1° and a sidelobe suppression ratio exceeding 20dB. For example, an antenna unit receives parameters {phase offset -1.46 radians, amplitude weight 0.601}, calibrates the delay of its transmitted waveform to 0.23 nanoseconds, and adjusts the power to 60.1% of the standard value. It collaborates with adjacent units to construct a highly focused detection beam with an azimuth of 310.5° and an elevation of 19.8°.

[0085] Finally, step 1033 is performed according to the pre-divided gridded target area, for example, the monitoring area is decomposed into a sequence of 15m×15m sub-areas, and the spatial coordinates of the center point of each sub-area are output to the combination of phase offset and signal amplitude weight in the order of the scanning path; the combination of phase offset and signal amplitude weight calculates a new beam pointing vector based on the coordinates, triggering the phase offset of the full array antenna unit, for example, an edge unit is updated from -1.2rad to +0.8rad and the signal amplitude weight, for example, the center unit weight is adjusted from 0.9 to 1.0 and recalculated; the new parameter group is synchronously loaded into the phased array through the high-speed bus within 20μs, driving step 1032 to generate a beam pointing strictly to the current sub-area. Directional beam (azimuth / elevation control accuracy ±0.1°); after the beam is stably locked, the array transmits a spread-spectrum frequency-modulated detection signal with a pulse width of 1ms, for example, at a center frequency of 5.8GHz. The transmit power is dynamically adjusted between 200-450W based on the amplitude weight. Simultaneously, the receiver collects reflected echoes, acquiring a 16,384-point IQ data stream through bandpass filtering and ADC sampling, while simultaneously recording delay information with 0.1ns accuracy. After completing single-point acquisition, immediate quality verification is performed, such as an 18dB signal-to-noise ratio threshold and a 0.3ns delay fluctuation threshold. If any failure occurs, the power is increased by 15% and a rescan is performed. Finally, a structured echo data packet with spatial coordinate markers is output and automatically iterates to the next sub-area until full coverage is achieved.

[0086] For example, an intelligent pod carried by a high-altitude platform performs monitoring tasks in an area with strong electromagnetic interference.

[0087] Specifically, the system first calculates a quantitative attenuation coefficient (0.85) based on meteorological data. Using the center of the monitoring area as the initial reference direction, the base phase offset of each cell is calculated using the location coordinates. This is then combined with the attenuation coefficient to generate a phase correction factor (e.g., +12°). Simultaneously, base amplitude weights are assigned based on the annular array layout (e.g., 0.9 for the center cell and 0.6 for the edge cells). The resulting beam is then enhanced by an amplitude scaling factor (1.2). The beam then scans along a raster path, emitting a 78 GHz frequency-modulated signal when directed toward sub-area A. The receiver then uses pulse compression processing to obtain high-resolution echo data from that area.

[0088] This step achieves optimized compensation of beam parameters by dynamically fusing the environmental attenuation coefficient and array geometric characteristics, significantly improving the signal-to-noise ratio and resolution of target echoes in complex environments. At the same time, it quickly switches the beam direction based on the electronic scanning mechanism to achieve high-efficiency and high-precision detection of wide-area targets, effectively suppressing sidelobe interference and enhancing the system's ability to resist environmental disturbances.

[0089] 104. Perform joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result.

[0090] Optionally, performing a joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result in step 104 includes:

[0091] 1041. Combining the radiation data stream and the target echo data into a multi-dimensional data set; and constructing a common representation space using a set of basis vectors;

[0092] 1042. Solve the sparse coefficient vector of the multidimensional data set in the common representation space, wherein the sparse coefficient vector satisfies the minimization of data reconstruction error and sparsity constraints;

[0093] The process of step 1042, "solving the sparse coefficient vector of the multidimensional data set in the common representation space," includes: constructing an optimization objective based on the multidimensional data set and a set of basis vectors in the common representation space, the optimization objective including a reconstruction error term and a sparsity constraint term, wherein the reconstruction error term represents the difference between the multidimensional data set and a linear combination of the basis vectors, and the sparsity constraint term represents the control of the number of non-zero elements of the sparse coefficient vector; solving the optimization objective through a preset iterative process to determine the sparse coefficient vector of the multidimensional data set;

[0094] The preset iterative process includes: initializing the sparse coefficient vector as a starting vector; in each iteration, updating each element of the sparse coefficient vector in turn, wherein the update of each element is based on the joint influence of the current reconstruction error and the sparsity constraint term, and checking whether the updated sparse coefficient vector meets the preset convergence condition, if so, terminating the iteration, otherwise continuing the iteration;

[0095] When each element of the sparse coefficient vector is updated, an intermediate value is calculated based on the basis vector and the multidimensional data set, and the intermediate value is compared with a regularization parameter to determine the final value of the element; the regularization parameter is used to balance the weights of the reconstruction error term and the sparsity constraint term, and remains unchanged during the iterative optimization process; the preset convergence condition is based on whether the change in the sparse coefficient vector in adjacent iterations is lower than a fixed threshold.

[0096] 1043. Reconstruct a fused data sequence representation based on the sparse coefficient vector as a multi-sensor data fusion result.

[0097] Among them, the process of step 1043 "reconstructing a fused data sequence representation based on the sparse coefficient vector as a multi-sensor data fusion result" includes: according to the sparse coefficient vector and a predefined basis vector set, multiplying each basis vector by the corresponding coefficient element in the sparse coefficient vector to generate a set of weighted basis vectors; superimposing all the weighted basis vectors item by item according to the element position to form a fused data sequence; dividing the fused data sequence according to the characteristic representation length of the radiation data stream and the target echo data, and outputting the multi-sensor data fusion result.

[0098] In the above scheme, the radiation data stream refers to time-series data consisting of target thermal radiation or spectral characteristics collected by infrared / optical sensors. Target echo data refers to the target reflection signal received by the phased array radar, containing information such as range, velocity, and scattering characteristics. The multidimensional data set is a matrix formed by stacking the radiation data stream and target echo data in time, with rows corresponding to sensor channels and columns corresponding to sampling times. The common representation space is a low-dimensional space spanned by a set of basis vectors, used to uniformly represent multi-source data. The sparse coefficient vector is the projection coefficient of the multidimensional data onto the basis vectors, with most elements being close to zero. The reconstruction error term is the mean squared error between the reconstructed data and the original data. The sparsity constraint term is a constraint on the number of nonzero elements in the coefficient vector. The regularization parameter is a hyperparameter that balances the reconstruction error and the strength of sparsity.

[0099] In the embodiment of the present application, first, step 1401 is used to strictly align the radiation data stream, i.e., the visible light / near infrared / thermal infrared three-band time series data, and the target echo data according to the millisecond timestamp. The spectral radiation value and the echo complex signal are uniformly encapsulated into a multidimensional data matrix through data type conversion, wherein the row vectors correspond to different sensor channels, for example, rows 1-3 are the three spectral bands, row 4 is the echo delay, row 5 is the echo real part, and row 6 is the echo imaginary part. The column vectors represent the time series sampling points, for example, 1000 columns correspond to a 1-second data window. Then, a core basis vector group of the common representation space is designed, that is, a wavelet basis is used for the spectral band to capture the sudden change information of the ground feature, such as a sudden drop in the vegetation index, and a Fourier basis is used for the echo data to analyze the periodic scattering characteristics, such as the vibration frequency of the equipment. A cross-domain hybrid orthogonal basis group is constructed through tensor product, and finally a common representation space architecture is formed that can simultaneously express the spectral radiation intensity distribution and the electromagnetic scattering characteristics. For example, the basis vector Φ_25 describes the coupling characteristics of the vegetation reflectivity and the specific resonant frequency.

[0100] Secondly, the all-zero sparse coefficient vector is first initialized as the starting point through step 1402; each coefficient element is updated sequentially in each iteration, that is, an intermediate value is generated by performing the inner product operation of the basis vector and the current data reconstruction residual, that is, the contribution of the basis vector to the residual is represented, and then the sparsity constraint is applied to compare the intermediate value with the preset regularization parameter, wherein if the absolute value is less than the parameter, it is forced to zero, otherwise its absolute value is reduced toward zero; this update process takes into account minimizing the reconstruction error and requires that the linear combination of the basis vectors closely fits the original data and the number of non-zero coefficients constrained by the sparsity control; after completing a single round of full-element update, the system calculates the change in the Euclidean distance between the current coefficient vector and the previous iteration result, and convergence is determined when the change in three consecutive iterations is less than a fixed threshold of 1e-4; finally, a sparse coefficient vector with a non-zero element ratio less than one thousandth is output, for example, a 1024-dimensional vector retains only 7 valid elements.

[0101] Finally, in step 1403, based on the non-zero elements in the sparse coefficient vector and their corresponding coefficient values, associated basis vectors are extracted from the common representation space. Each basis vector is then multiplied by the coefficient to perform amplitude weighting. A fused data sequence of the original data dimension is reconstructed through a linear superposition operation of the weighted basis vectors, where the superposition process strictly follows the positioning index of each basis vector in the spatiotemporal coordinate system. The reconstructed sequence is then divided according to the sensor feature length: the front section is restored to the spectral radiation characteristics, that is, the visible light / near infrared / thermal infrared three-band time series data is separated, and the back section is restored to the target echo characteristics, that is, the delay mark and IQ signal stream are reconstructed. Finally, the multi-sensor data fusion result is encapsulated as a multidimensional data matrix with a unified timestamp and output. Its row and column structure remains consistent with the input data, but the key features are significantly enhanced in the sparse representation. For example, the reflectivity mutation in the vegetation stress area and the equipment vibration echo frequency form a spatiotemporally aligned fusion feature, which directly supports the multi-dimensional joint analysis and decision-making of the target.

[0102] For example, a security system synchronously obtains the infrared radiation sequence and radar echo sequence of the target area.

[0103] Specifically, the two are first aligned in time to form a 2x20-dimensional data matrix. A DCT basis is then used to construct a 100-dimensional common representation space. Sparse coefficients are solved using coordinate descent, for example, with a regularization parameter of λ = 0.1, and convergence is achieved after 15 iterations. During reconstruction, key basis vectors are extracted, such as the DCT basis vectors 1, 5, and 7, and weighted superposition is performed to generate a fused 20-frame enhanced feature sequence, in which the infrared hot spots and radar scatter points are associated as the same target.

[0104] This step effectively suppresses sensor noise and redundant information by mining the common features of multi-source data through joint sparse constraints; the fusion result retains the complementary advantages of radiation and scattering characteristics, improves the significance and separability of target features, and provides a high-value data foundation for subsequent identification.

[0105] The following is a complete example of steps 101 to 104:

[0106] During an intelligent inspection mission in the high-voltage transmission corridor of City D, a smart pod equipped with a multispectral sensor and phased array antenna flew along the transmission line, performing the following continuous operations:

[0107] First, when the pod entered a strong electromagnetic interference area, 50 meters from a 500 kV line, the multispectral sensor simultaneously captured three-band radiation data from the insulator string below. The visible light channel identified contaminated textures on the porcelain bottle surface, the near-infrared channel analyzed the reflective characteristics of the composite insulating silicone, and the thermal infrared channel detected the abnormal temperature rise point of insulator #07. The three-channel data was time-aligned to generate a continuous radiation data stream with a 10 Hz sampling rate.

[0108] Secondly, based on the radiation transfer model, the normal radiation value of position #07 is 850, which is compared with the measured value of only 510 due to interference. The attenuation ratio of the thermal infrared band is calculated to be 0.385. After normalization, the output quantitative attenuation coefficient is 98 (integer value 0-255), and the visible light / near infrared coefficient is obtained at the same time [35,72].

[0109] Then, a quantized attenuation coefficient of 98 was used to drive phased array beam control. Specifically, the insulator cluster was divided into a 2×2 grid, and the beam parameters for the G02 grid (including the #07 insulator) were calculated. The phase offset of the edge cells was updated to +1.2 rad (to compensate for 30% phase distortion), and the amplitude weight of the center cell was reduced to 0.6 (to prevent signal saturation). A highly directional beam with an azimuth of 283°±0.1° was generated within 20 μs. A 1 ms frequency-modulated pulse (5.8 GHz) was transmitted, and the echo was received. The scattering point at the #07 steel cap was precisely located using a 32.4 ns delay, and an echo data packet with coordinates was output.

[0110] Finally, the radiation data stream (including the 64°C hot spot) and the G02 echo data (32.4ns scattering characteristics) were aligned in time and space, and the sparse representation was achieved through a Fourier mixed basis. That is, a 0.3% sparsity coefficient vector was obtained after 1024 iterations. The fusion sequence was reconstructed to synchronously enhance the thermal anomaly and defect scattering characteristics, and the diagnostic conclusion of "#07 insulator steel cap corrosion with local overheating" was directly output, verifying that the fusion method can accurately identify complex defects missed by traditional single sensors in a strong interference environment.

[0111] Figure 2 The present invention provides a structural diagram of a multi-sensor data fusion system based on an intelligent pod, as shown in FIG. Figure 2 As shown, the system includes:

[0112] The first generation module 21 collects the radiation data stream generated by the multispectral sensor of the smart pod in the electromagnetic interference field, wherein the radiation data stream includes radiation data in the visible light band, the near infrared band, and the thermal infrared band;

[0113] A calculation module 22 calculates the quantitative attenuation coefficient of the radiation value of each band in the electromagnetic interference field using a radiation transfer equation based on the radiation data stream;

[0114] A second generating module 23 generates a directional beam corresponding to the phased array antenna array in the intelligent pod based on the quantized attenuation coefficient, and uses the directional beam to scan the target area to obtain target echo data;

[0115] The third generating module 24 performs a joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result.

[0116] Figure 2 The multi-sensor data fusion system based on the intelligent pod can perform Figure 1 The implementation principles and technical effects of the multi-sensor data fusion method based on the smart pod described in the illustrated embodiment will not be elaborated on here. The specific manner in which the various modules and units perform operations in the multi-sensor data fusion system based on the smart pod in the above embodiment have been described in detail in the relevant embodiments of the method and will not be elaborated on here.

[0117] In one possible design, Figure 2 The multi-sensor data fusion system based on the smart pod of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0118] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0119] The processing component 32 is used for the above Figure 1 The embodiment of the invention provides a multi-sensor data fusion method based on an intelligent pod.

[0120] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0121] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0122] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0123] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0124] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0125] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0126] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The multi-sensor data fusion method based on the smart pod of the illustrated embodiment.

[0127] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0128] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0129] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-sensor data fusion method based on an intelligent pod, characterized in that: include: Collecting radiation data streams generated by the multispectral sensor of the smart pod in an electromagnetic interference field, wherein the radiation data streams include radiation data in the visible light band, near infrared band, and thermal infrared band; Based on the radiation data stream, calculating the quantitative attenuation coefficient of the radiation value of each band in the electromagnetic interference field by using a radiation transfer equation; Based on the quantized attenuation coefficient, a directional beam corresponding to the phased array antenna array in the intelligent pod is generated, and the directional beam is used to scan the target area to obtain target echo data; Performing a joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result; Generating a directional beam corresponding to the phased array antenna array in the intelligent pod based on the quantized attenuation coefficient, and scanning a target area using the directional beam to obtain target echo data, including: Determining adjustment parameters of each antenna element of a phased array antenna array in the smart pod based on the quantized attenuation coefficient, wherein the adjustment parameters include a phase offset and a signal amplitude weight; According to the adjustment parameters, corresponding phase offsets and signal amplitude weights are simultaneously applied to all antenna elements of the phased array antenna array to generate a directional beam corresponding to the phased array antenna array in the smart pod; The combination of phase offset and signal amplitude weight is repeatedly changed so that the directional beam is directed to different sub-areas of the target area in turn. In the direction of each sub-area, a detection signal is transmitted through the phased array antenna array, and a reflected signal from the target area is received as target echo data.

2. The method according to claim 1, characterized in that Performing a joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result, including: Combining the radiation data stream and the target echo data into a multi-dimensional data set; and constructing a common representation space using a set of basis vectors; In the common representation space, solving a sparse coefficient vector of the multidimensional data set, wherein the sparse coefficient vector satisfies a minimization of data reconstruction error and a sparsity constraint; Based on the sparse coefficient vector, a fused data sequence representation is reconstructed as a multi-sensor data fusion result.

3. The method according to claim 1, characterized in that Based on the quantized attenuation coefficient, determining adjustment parameters of each antenna element of the phased array antenna array in the smart pod, wherein the adjustment parameters include a phase offset and a signal amplitude weight, including: Obtaining the physical position of each antenna unit in the phased array antenna array as a position coordinate; An initial phase value is obtained by performing a dot product operation on the position coordinates of the antenna unit and a reference direction vector, and then the initial phase value is multiplied by a constant factor related to the signal wavelength to obtain a basic phase offset; Inputting the basic phase offset into a predefined phase adjustment function, using the phase adjustment function to take the quantized attenuation coefficient as an input variable, outputting a phase correction factor, and multiplying the basic phase offset by the phase correction factor to obtain a phase offset; Assigning a weight related to the distance from the array center to each antenna element based on the relative position of the antenna element in the phased array antenna array using a predefined position weight function to obtain a basic signal amplitude weight; The basic signal amplitude weight is input into a predefined amplitude adjustment function, and the amplitude adjustment function outputs an amplitude scaling factor with the quantized attenuation coefficient as an input variable, and the basic signal amplitude weight is multiplied by the amplitude scaling factor to obtain the signal amplitude weight.

4. The method according to claim 2, characterized in that Solving the sparse coefficient vector of the multidimensional data set in the common representation space includes: constructing an optimization objective based on the multidimensional data set and a set of basis vectors in the common representation space, the optimization objective comprising a reconstruction error term and a sparsity constraint term, wherein the reconstruction error term represents a difference between the multidimensional data set and a linear combination of the basis vectors, and the sparsity constraint term represents a control of the number of non-zero elements of the sparse coefficient vector; solving the optimization objective through a preset iterative process to determine a sparse coefficient vector of the multidimensional data set; The preset iterative process includes: initializing the sparse coefficient vector as a starting vector; in each iteration, updating each element of the sparse coefficient vector in turn, wherein the update of each element is based on the joint influence of the current reconstruction error and the sparsity constraint term, and checking whether the updated sparse coefficient vector meets the preset convergence condition, if so, terminating the iteration, otherwise continuing the iteration; When each element of the sparse coefficient vector is updated, an intermediate value is calculated based on the basis vector and the multidimensional data set, and the intermediate value is compared with a regularization parameter to determine the final value of the element; the regularization parameter is used to balance the weights of the reconstruction error term and the sparsity constraint term, and remains unchanged during the iterative optimization process; the preset convergence condition is based on whether the change in the sparse coefficient vector in adjacent iterations is lower than a fixed threshold.

5. The method according to claim 2, characterized in that Based on the sparse coefficient vector, a fused data sequence representation is reconstructed as a multi-sensor data fusion result, including: According to the sparse coefficient vector and a predefined set of basis vectors, multiplying each basis vector by a corresponding coefficient element in the sparse coefficient vector to generate a set of weighted basis vectors; Superimposing all the weighted basis vectors one by one according to element positions to form a fused data sequence; The fused data sequence is divided according to the characteristic representation lengths of the radiation data stream and the target echo data, and a multi-sensor data fusion result is output.

6. The method according to claim 1, wherein Based on the radiation data stream, the quantitative attenuation coefficient of the radiation value of each band in the electromagnetic interference field is calculated by using a radiation transfer equation, including: For each spectral band in the radiation data stream, extracting a reference radiation value of the spectral band in a field without electromagnetic interference; Inputting the reference radiation value and the actual radiation value under the electromagnetic interference field into the radiation transfer equation to calculate the radiation attenuation ratio of each band; The radiation attenuation ratio is numerically quantified and the output is a quantized attenuation coefficient.

7. A multi-sensor data fusion system based on an intelligent pod, used in the multi-sensor data fusion method based on an intelligent pod according to any one of claims 1 to 6, characterized in that: include: The first generation module collects the radiation data stream generated by the multispectral sensor of the intelligent pod in the electromagnetic interference field, and the radiation data stream includes radiation data in the visible light band, the near infrared band, and the thermal infrared band; A calculation module, which calculates the quantitative attenuation coefficient of the radiation value of each band in the electromagnetic interference field through a radiation transfer equation based on the radiation data stream; A second generating module generates a directional beam corresponding to the phased array antenna array in the intelligent pod based on the quantized attenuation coefficient, and uses the directional beam to scan the target area to obtain target echo data; The third generation module performs a joint sparse representation on the radiation data stream and the target echo data to generate a multi-sensor data fusion result.

8. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the multi-sensor data fusion method based on the smart pod as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the multi-sensor data fusion method based on the smart pod according to any one of claims 1 to 6 is implemented.

Images