A non-contact electromagnetic detection system based on 3DFWI algorithm and permittivity inversion algorithm

Abstract

This invention discloses a non-contact electromagnetic detection system based on a 3D FWI algorithm and a dielectric constant inversion algorithm, relating to the fields of non-destructive testing and electromagnetic imaging technology. It includes a multi-transmitter module array, a multi-channel receiver module array, a timing control and electronic scanning module, a data acquisition and synchronization system, a signal processing unit, and a 3D imaging and visualization terminal. This invention employs a fixed multi-Tx-Rx antenna array and electronic time-division multiplexing scanning technology to achieve high-speed data acquisition; and integrates a 3D full-wave inversion algorithm and a dielectric constant inversion algorithm to directly reconstruct the 3D dielectric constant distribution within the tested medium. This invention solves the problems of contact coupling, low detection efficiency, insufficient resolution, and inability to quantitatively image 3D in existing non-destructive testing technologies such as ultrasonic, resistivity, and traditional ground-penetrating radar. It achieves high-precision, real-time, non-contact 3D visualization and quantitative analysis of the internal structures of concrete, composite materials, etc.

Description

Technical Field

[0001] This invention relates to the fields of non-destructive testing, electromagnetic imaging technology and algorithms, and in particular to a non-contact electromagnetic testing system based on the 3DFWI algorithm and dielectric constant inversion algorithm. Background Technology

[0002] Non-destructive testing (NDT) technology is crucial for assessing the internal condition of engineering structures such as concrete, composite materials, and geotechnical materials, and is a core means of ensuring infrastructure safety. Currently, the technologies widely relied upon in engineering practice mainly include ultrasonic testing, eddy current testing, resistivity methods, and radiographic testing. However, these traditional methods have inherent limitations in their physical principles, making it difficult for them to meet the growing demands of modern engineering for high-precision, high-efficiency, and non-destructive testing in terms of application scenarios, accuracy, and efficiency. For example, acoustic testing is susceptible to the influence of coupling states and metal interfaces; electrical methods struggle to achieve precise quantification; and radiographic testing faces the dual challenges of cost and safety.

[0003] Specifically, existing mainstream technologies face severe challenges: First, while ultrasonic testing is relatively sensitive to homogeneous defects within concrete, its sound wave energy is strongly reflected at the interface between the steel reinforcement and concrete, making it difficult to effectively detect defects behind or around metal components. Furthermore, it has stringent requirements for the flatness of the test surface and the coupling agent, resulting in low testing efficiency. Second, the ground resistivity method based on electromagnetic principles reflects the macroscopic electrical parameters of the medium, only providing a vague qualitative identification of abnormal areas. It cannot provide precise geometric morphology, size, and three-dimensional spatial location of defects, significantly limiting its practicality. Third, although traditional contact ground-penetrating radar can utilize electromagnetic waves for detection, its single-point or line scanning mode heavily relies on mechanical movement, resulting in slow data acquisition speed. Simultaneously, in media with high water content or high conductivity, electromagnetic wave energy attenuates sharply, leading to a significant decrease in effective detection depth and imaging resolution.

[0004] In recent years, researchers both domestically and internationally have attempted to combine advanced algorithms such as machine learning with traditional detection methods to improve detection capabilities. However, these emerging solutions still face numerous limitations in complex engineering environments: they either rely on large amounts of high-quality, labeled training data (which is difficult to obtain in practice) or fail to fundamentally overcome the bottlenecks in sensor hardware and data acquisition methods. Therefore, the current technological landscape is fragmented: most methods only provide two-dimensional information; some technologies capable of three-dimensional imaging are difficult to popularize due to cumbersome processing procedures, excessive time consumption, or high equipment costs; solving a complex problem often requires the coordinated use of multiple detection methods, leading to complex processes and a large workload. The engineering field urgently needs a revolutionary detection technology that integrates non-contact, high-resolution, true 3D imaging and real-time on-site analysis. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm. Through in-depth optimization of the antenna array, scanning strategy and core inversion algorithm, it can achieve detection targets that do not require coupling medium, have centimeter-level spatial resolution, and can perform quantitative analysis and real-time three-dimensional visualization of internal defects in complex media.

[0006] This invention discloses a non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm, comprising: a multi-transmitter module, consisting of multiple transmitting antennas, for generating and radiating ultra-wideband pulsed electromagnetic waves; a multi-channel receiver module, consisting of multiple receiving antennas, forming a fixed antenna array together with the multi-transmitter module; a timing control and electronic scanning module, configured with time-division multiplexing control logic, for sequentially energizing each transmitting antenna in the multi-transmitter module and synchronously triggering the multi-channel receiver module to perform data acquisition; a data acquisition and synchronization system, for synchronously acquiring and recording the echo signals received by the multi-channel receiver module; a signal processing unit, with a built-in hardware acceleration computing module; and a three-dimensional imaging and visualization terminal; wherein, the signal processing unit is configured to: execute the three-dimensional full-wave inversion algorithm, process the acquired echo data, and invert the three-dimensional dielectric constant distribution model inside the target under test.

[0007] This invention deeply integrates a fixed multi-antenna array with a three-dimensional full-wave inversion algorithm, achieving a fundamental leap from traditional "indirect inference" or "two-dimensional imaging" to direct and quantitative inversion of the three-dimensional dielectric constant distribution inside the medium. This provides unprecedented physical accuracy and rich three-dimensional information dimensions for non-destructive testing.

[0008] To optimize the technical solution of the above system, the multi-transmitter module and the multi-channel receiver module are integrated into a fixed, non-mechanically moving transceiver probe. This solution eliminates the problems of bulkiness, slow speed, and susceptibility to failure associated with mechanical scanning mechanisms by integrating the multi-transmitter and multi-channel receiver modules into a fixed, integrated probe. This improves the system's integration, reliability, and data acquisition efficiency, making it highly suitable for deployment in space-constrained or rapid-operation environments.

[0009] To optimize the technical solution of the aforementioned system, the timing control and electronic scanning module achieves nanosecond-level switching of the transmitting antennas through a high-speed RF switch array, ensuring that only one transmitting antenna in the multi-transmitting module is activated at any given time. Simultaneously, all receiving antennas in the multi-channel receiving module acquire data in parallel. This timing control and time-division multiplexing excitation and acquisition strategy fundamentally eliminates the instantaneous electromagnetic crosstalk path between multiple transmitting antennas. This effectively suppresses the main noise at the system source, significantly improving the signal-to-noise ratio and purity of the original acquired data, laying a solid data foundation for subsequent high-precision inversion.

[0010] To optimize the technical solution of the above system, the multi-transmission module is configured to switch transmission signals between multiple preset center frequency points; the signal processing unit is further configured to perform spectral separation and weighted fusion processing on the data collected at different center frequencies. This solution, by configuring multiple transmission modules for multi-band switching transmission and combining it with frequency domain fusion processing, can utilize the complementarity of different frequency electromagnetic waves in defect detection (low frequency has deep penetration, high frequency has high resolution), while effectively suppressing narrowband interference in specific environments through weighted fusion, thereby achieving a better balance between detection depth and imaging quality in complex media.

[0011] To optimize the technical solution of the above system, the hardware acceleration computing module is a GPU, and the 3D full-wave inversion algorithm is accelerated using the parallel computing architecture of the GPU. Here, the calculated 3D FWI algorithm is deployed on a GPU parallel computing architecture, compressing the inversion calculation process, which originally required hours or even days, to minutes. This achieves a revolutionary shift from "post-processing" to "real-time on-site imaging," improving detection efficiency and engineering practical value.

[0012] To optimize the technical solution of the aforementioned system, the three-dimensional full-wave inversion algorithm employs a multi-scale or multi-frequency iterative strategy to reduce its dependence on the initial model. This approach enhances the robustness of the algorithm by using a multi-scale or multi-frequency iterative inversion strategy, ensuring that the accuracy of the final inversion result no longer excessively depends on the precision of the initial model. This reduces the algorithm's reliance on prior knowledge and improves the system's adaptability and stability in unknown and complex environments.

[0013] To optimize the technical solution of the aforementioned system, the data acquisition and synchronization system employs a high-precision phase-locked loop or atomic clock as the synchronization source to ensure strict time synchronization between transmission and acquisition events. This high-precision clock synchronization mechanism ensures extremely high phase consistency of massive channel data on the time axis, a prerequisite for achieving high-precision electron synthetic aperture and reliable waveform comparison, directly determining the spatial positioning accuracy and geometric fidelity of the 3D reconstruction model.

[0014] To optimize the technical solution of the aforementioned system, the 3D imaging and visualization terminal is configured to: display the voxelized 3D dielectric constant distribution model obtained from the inversion, and automatically and quantitatively calculate the geometric parameters of internal defects through image segmentation and feature extraction algorithms. This solution, by integrating automated image segmentation and quantitative calculation functions into the terminal, transforms the abstract dielectric constant distribution map into key geometric parameters that engineers can directly understand, achieving an upgrade from "qualitative imaging" to "quantitative diagnosis," and enhancing the intuitiveness and decision support value of the detection results.

[0015] This invention also introduces a non-contact electromagnetic detection method using the above system, comprising the following steps: 1) Arranging the multi-transmitter module and multi-channel receiver module in a non-contact manner above the target under test; 2) Exciting each transmitting antenna sequentially according to the timing control and electronic scanning module, and synchronously acquiring the echo signals of all receiving antennas during each excitation; 3) Preprocessing the acquired multi-channel echo signals; 4) Inputting the preprocessed data into the signal processing unit, executing the GPU-accelerated three-dimensional full-wave inversion algorithm, and inverting to obtain the three-dimensional dielectric constant distribution model inside the target under test; 5) Visualizing and quantitatively analyzing the three-dimensional dielectric constant distribution model in the three-dimensional imaging and visualization terminal.

[0016] The detection method of this solution provides a standardized and efficient operating paradigm, ensuring that the entire process from non-contact data acquisition to real-time 3D visualization can be executed quickly and reliably, and guaranteeing the repeatability and reproducibility of the excellent performance of this invention in practical applications.

[0017] To further optimize the aforementioned non-contact electromagnetic detection method, a velocity calibration step is included before executing the three-dimensional full-wave inversion algorithm: the propagation velocity of electromagnetic waves in the measured medium is calibrated on-site using a pre-set calibration target. This solution can perform adaptive calibration for different medium conditions, effectively compensating for measurement errors caused by the uncertainty of the electromagnetic wave velocity in the medium, thus ensuring the long-term accuracy of quantitative measurement results under different detection environments.

[0018] Compared with the prior art, the beneficial effects of the present invention are: 1. In terms of detection performance, this system achieves a balance between high precision, quantification, and strong adaptability. The three-dimensional full-wave inversion (3D FWI) algorithm used in this system fully utilizes the full waveform information of electromagnetic waves (including amplitude, phase, and travel time), which can accurately characterize the scattering, refraction, and diffraction behavior of electromagnetic waves in non-uniform media. This enables the system to not only achieve clear imaging with spatial resolution at the centimeter or even millimeter level, but also to directly output a three-dimensional distribution map of dielectric constant with clear physical meaning, thereby achieving accurate quantification of key parameters such as rebar diameter, protective layer thickness, void volume, and crack aperture. At the same time, the multi-band fusion strategy and advanced noise reduction algorithm effectively overcome the technical challenge of weak signals in high-attenuation media, significantly improving the system's detection capability and imaging robustness under complex working conditions such as wet concrete and high-salt environments.

[0019] 2. This invention represents a revolutionary improvement in detection efficiency and ease of use, truly meeting the needs of real-time on-site diagnosis. Through innovative fixed multi-Tx-Rx arrays and electronic time-division multiplexing (TDM) scanning technology, it eliminates the bulky and slow mechanical scanning mechanism. Data acquisition can be completed in just seconds to minutes, increasing detection efficiency by tens of times compared to traditional methods. Combined with a built-in FPGA+GPU heterogeneous parallel computing architecture, the three-dimensional full-wave inversion calculation process, which originally took hours or even days on large servers, is compressed to a "minute-level" or even "near real-time" level acceptable for engineering sites. Operationally, the non-contact measurement method requires no coupling agent and has no strict requirements on the condition of the measured surface, reducing operational difficulty and workload, making rapid, large-scale surveys possible.

[0020] 3. This system demonstrates outstanding comprehensive advantages and universal value in terms of system integration and application breadth. It highly integrates high-speed electronic scanning, massive data acquisition, and real-time 3D inversion functions into a compact, portable, and fully functional solution. This highly integrated design not only reduces the system's dependence on external motion control precision and enhances overall stability and reliability, but also makes it extremely flexible in deployment, making it ideal for use in space-constrained field applications such as under bridges and inside tunnels. Therefore, its application scope can widely cover multiple fields such as civil engineering, transportation infrastructure, and materials science. It can provide a universal and highly reliable internal non-destructive testing and condition assessment method for diverse testing objects such as reinforced concrete structures, composite material components, and soil and rock masses, possessing enormous promotional value and application prospects. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm of the present invention.

[0022] In the diagram, 1 is a multi-transmit module; 2 is a multi-channel receiver module; 3 is a timing control and electronic scanning module; 4 is a data acquisition and synchronization system; 5 is a signal processing unit; and 6 is a 3D imaging and visualization terminal. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0024] This invention relates to a non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm. Its core lies in the combination of innovative hardware architecture and advanced software algorithms to achieve three-dimensional, quantitative, and visualized non-destructive testing of the internal structure of non-homogeneous media such as concrete and composite materials. The system emits ultra-wideband electromagnetic waves and receives echo signals reflected from the internal structure. Finally, it uses the three-dimensional full-wave inversion algorithm to reconstruct the three-dimensional distribution model of the dielectric constant inside the medium, thereby accurately identifying targets such as reinforcing bars, cracks, voids, and water-bearing areas.

[0025] like Figure 1 As shown, the present invention includes a multi-transmit module 1 composed of multiple ultra-wideband transmit antennas and a multi-channel receive module 2 composed of multiple high-sensitivity receive antennas. These antenna elements are precisely arranged according to an optimized planar matrix topology to form a compact multiple-input multiple-output (MIMO) array. Specifically, the transmit and receive antennas are arranged in an interleaved manner, for example, combining a 4×4 transmit antenna array with an 8×8 receive antenna array, achieving equivalent virtual array sampling within a limited physical aperture.

[0026] Each antenna element is specially designed to operate in the ultra-wideband frequency band and uses air coupling, eliminating the need for direct contact with the surface being measured. The antenna array is encapsulated in a protective shell made of high-strength composite material, which integrates miniaturized RF front-end circuitry and a temperature compensation system to ensure stable electrical performance under various environmental conditions. This integrated design makes the probe compact and lightweight, suitable for handheld operation or mounted on a robotic arm or mobile platform for automated scanning.

[0027] The system of this invention adopts a highly integrated modular architecture, and its core modules achieve fully automated detection through precise collaboration. The system's command center—the timing control and electronic scanning module 3—sends nanosecond-precision excitation control signals to the multi-transmission module 1 via a high-speed control bus, and simultaneously sends a global acquisition trigger signal to the data acquisition and synchronization system 4 via a synchronous trigger line. After the multi-transmission module 1 radiates ultra-wideband pulsed electromagnetic waves to the target under test according to the instructions, the multi-channel receiving module 2 synchronously captures the analog echo signal reflected by the target and transmits the signal to the data acquisition and synchronization system 4 through multiple independent radio frequency channels. Upon receiving the trigger signal, the data acquisition system 4 immediately activates all ADC channels, converting the analog signal into digital waveform data, and transmits it via a PC. High-speed interfaces such as Ie transmit massive amounts of data to the signal processing unit 5. The signal processing unit 5, relying on a GPU-accelerated architecture, preprocesses the data and performs three-dimensional full-wave inversion calculations. After generating three-dimensional dielectric constant volume data, the processing results are transmitted to the three-dimensional imaging and visualization terminal 6 through a graphics data interface. The three-dimensional imaging and visualization terminal 6 renders interactive three-dimensional images in real time and generates quantitative diagnostic reports. At the same time, it receives user commands and realizes closed-loop control of the entire system through the processing unit 5. This tightly coupled signal transmission architecture ensures synchronization and precise coordination throughout the entire process from electromagnetic wave transmission, data acquisition, real-time processing to three-dimensional imaging.

[0028] The multi-transmitter module 1 adopts a software-defined radio frequency architecture, enabling intelligent switching between multiple preset center frequencies. The system is configured with at least three characteristic frequency bands: Low-frequency detection band: It has strong penetration capability and can maintain effective detection depth even in attenuating media such as water-containing concrete. It is suitable for the preliminary identification of deep defects and large-scale targets.

[0029] Mid-frequency imaging band: Achieving the best balance between penetration depth and resolution, it is the main imaging band and provides core data for most defect detections.

[0030] High-frequency fine segment: provides extremely high spatial resolution and has unique advantages in detecting fine structures such as microcracks and protective layer thickness.

[0031] This invention completely abandons traditional mechanical motion mechanisms by employing a fixed array, replacing them with a purely electronic scanning method. Specifically, through precise control of timing and the electronic scanning module 3, the system uses a time-division multiplexing (TDM) mechanism to sequentially activate each transmitting antenna at nanosecond-level switching speeds. When a transmitting antenna is activated, all receiving antennas synchronously acquire signals, thus obtaining a complete dataset from different spatial locations and angles in a single fixed-position measurement. This fixed multi-antenna array design is not only a component-level improvement but also a system-level innovation. Through hardware architecture innovation, it provides an unprecedented high-quality, high-density data foundation for subsequent three-dimensional full-wave inversion algorithms. Spatial sampling and diverse observation angles enhance the well-posedness of the inverse problem, making accurate reconstruction of minute defects in complex media possible. Simultaneously, this design simplifies on-site operation procedures; operators only need to place the probe above the area to be tested to begin inspection, reducing the technical requirements for operators and laying the foundation for the large-scale application of high-tech non-destructive testing technology.

[0032] The electronic scanning and synchronous acquisition system used in this invention is the core technical architecture for achieving high-performance detection. It revolutionizes the data acquisition mode of traditional electromagnetic detection through precise timing control and nanosecond-level synchronization technology.

[0033] The timing control and electronic scanning module 3 is a high-precision timing control system built on a programmable logic device (FPGA). The time division multiplexing (TDM) control logic built into the timing control and electronic scanning module 3 achieves precise polling excitation of each transmit antenna in the multi-transmit module 1 through a set of high-speed radio frequency switch arrays (switching time < 10 nanoseconds).

[0034] The working mechanism of the timing control and electronic scanning module 3 is as follows: 1) Preprogrammed sequence execution: The system activates each transmit antenna in the array sequentially according to a pre-optimized transmit sequence. For example, in an array containing 16 transmit antennas, the module generates 16 consecutive time slots, in which only one specified transmit antenna is activated.

[0035] 2) Nanosecond-level timing control: The start time, pulse width and repetition frequency of each transmitted pulse are precisely controlled by digital trigger signals generated by the FPGA, with a timing control accuracy better than 1 nanosecond. This extreme timing accuracy ensures the coherence of the phase of the transmitted electromagnetic wave, providing a key guarantee for subsequent synthetic aperture processing and waveform inversion.

[0036] 3) Intelligent scheduling algorithm: The TDM sequence has been specially optimized to avoid antennas in adjacent spatial locations from transmitting continuously in time, further reducing potential electromagnetic coupling interference and improving data quality.

[0037] This invention employs a unique "one-transmit, multiple-receive" parallel acquisition architecture, forming a highly efficient multiple-input multiple-output (MIMO) system: when any transmitting antenna (e.g., Txi) is excited and radiates an ultra-wideband pulse, all receiving antennas (e.g., 64) in the multi-channel receiving module 2 will start synchronously and acquire the scattered echo signals from the target area in parallel. Each receiving channel independently records the complete time-domain waveform, including amplitude, phase, and time of arrival, forming a slice of a data cube. After the system traverses all transmitting antennas, a complete MIMO dataset can be obtained. For a 16-transmit, 64-receive array, 16 × 64 = 1024 independent propagation path information will be obtained.

[0038] The present invention achieves the following technical advantages through timing control and electronic scanning module 3: 1) Fundamentally eliminate crosstalk: The strict TDM mechanism ensures that only one transmitter is working at any given time, completely solving the problem of mutual interference when multiple transmitters are working simultaneously.

[0039] 2) Data density and angle diversity: This design is equivalent to forming a virtual array in physical space, providing massive amounts of observation data from different incident and scattering angles, which enhances the constraints of subsequent three-dimensional inversion problems and is the key to achieving high-resolution reconstruction.

[0040] 3) Maximize efficiency: Compared with traditional single-transmit and single-receive point-by-point scanning, the data acquisition speed is increased by tens of times while maintaining an extremely high spatial sampling rate.

[0041] To ensure time consistency in large-scale array acquisition, this system constructs a sophisticated four-level synchronization architecture: 1. Master clock source: A temperature-compensated oven-controlled crystal oscillator (OCXO) or a miniaturized rubidium atomic clock is used as the primary clock source, with a frequency stability better than ±0.1 ppm, providing the system's time reference.

[0042] 2. Clock distribution network: The master clock signal is multiplied and distributed to all critical subsystems, including the transmitter clock, ADC sampling clock and FPGA control clock, through a phase-locked loop (PLL) circuit.

[0043] 3. Trigger Synchronization Mechanism: A unified global trigger signal generated by the FPGA is simultaneously sent to the transmitter and the ADCs of all receiving channels to ensure that the "transmission start time" and the "sampling start time" are strictly aligned.

[0044] 4. Phase synchronization calibration: During the initialization phase, the system will execute a self-calibration procedure to measure and compensate for the tiny time offset (on the order of picoseconds) between channels caused by wiring differences, thereby achieving true phase synchronization.

[0045] The innovative design of the electronic scanning and synchronous acquisition system in this invention enables the system to acquire the equivalent amount of data that would take traditional methods several hours to obtain in seconds, while ensuring that the data quality meets the stringent requirements of the most advanced three-dimensional full-wave inversion algorithm for data consistency and accuracy.

[0046] The data acquisition and synchronization system 4 of this system adopts a hierarchical synchronization architecture to achieve nanosecond-level time synchronization control. It uses a highly stable oven-controlled crystal oscillator or a miniaturized rubidium atomic clock as the primary master clock source, generating multiple reference frequencies required by the system through a phase-locked loop circuit, and distributing them to all acquisition channels via a precisely designed clock tree network. The system strictly aligns the transmission pulse and sampling timing through a unified global trigger signal—when the timing control and electronic scanning module 3 issues a transmission command, the synchronization trigger signal simultaneously activates the pulse generation circuit of the multi-transmission module 1 and all analog-to-digital converter channels of the data acquisition and synchronization system 4, ensuring strict time correlation throughout the entire link from signal transmission to acquisition. To achieve higher-precision synchronization, the system executes a channel delay calibration procedure during the initialization phase. By measuring the fixed transmission delay of each channel and performing digital compensation, the time synchronization error between all channels is ultimately controlled within 1 nanosecond, providing high-fidelity time-domain data for subsequent signal processing.

[0047] The sophisticated synchronization mechanism of the data acquisition and synchronization system 4 provides crucial assurance for system performance: First, it establishes a unified time reference for the massive amounts of data acquired by the multi-channel receiving module 2, enabling accurate coherent superposition and comparison of signals from different spatial locations—the foundation of synthetic aperture processing and beamforming technology. Second, strict time synchronization ensures the complete preservation of the high-precision phase information required by the full waveform inversion algorithm, and phase consistency is the core prerequisite for achieving millimeter-level spatial resolution and accurate dielectric constant inversion. The data acquisition and synchronization system 4 also possesses excellent environmental adaptability; its temperature compensation mechanism suppresses clock drift during operation, ensuring stable synchronization performance during long-term continuous detection, thus providing a reliable time reference for the entire 3D imaging system.

[0048] In this invention, the signal processing unit 5 achieves signal optimization and information fusion through a multi-level processing flow. First, it systematically preprocesses the acquired multi-frequency band data: joint time-frequency analysis using short-time Fourier transform and wavelet transform is performed to accurately extract the amplitude and phase characteristics of each frequency band signal; adaptive notch filtering for narrowband interference such as 50Hz power frequency and its harmonics, and civilian radio frequency bands is achieved by establishing an environmental electromagnetic interference spectrum template; waveform distortion of electromagnetic waves propagating in the medium is corrected based on dispersion relation analysis, effectively restoring the original signal characteristics. On this basis, the system adopts a three-level fusion architecture—feature-level fusion extracts characteristic parameters such as the maximum amplitude and instantaneous frequency of each frequency band, and generates an enhanced feature set based on the entropy weight method to determine weight coefficients; joint regularization inversion is implemented at the data level, and the optimal dielectric constant distribution is solved using the alternating direction multiplier method; wavelet decomposition and coefficient weighted fusion are used at the image level to finally obtain a fused image with rich details and suppressed noise.

[0049] The signal processing unit 5 significantly improves signal quality and imaging performance through multiple techniques, including coherent accumulation, incoherent processing, and spatial filtering. Specifically, it utilizes the coherent characteristics of multi-band signals to enhance phase alignment, suppresses random noise through multi-band joint estimation, and achieves spatial separation of the signal by combining array geometric information. The system's built-in quality assessment system can monitor key indicators such as signal-to-noise ratio and dynamic range in real time, evaluate the fusion effect based on information entropy theory, and dynamically optimize processing parameters through an adaptive feedback mechanism. These technological innovations increase the system's detection depth by approximately 40%, achieving spatial resolution at the centimeter (lateral) and millimeter (vertical) levels. It effectively suppresses imaging artifacts and maintains reliable detection capabilities even when some frequency bands are interfered with, ultimately achieving stable and accurate detection under various complex conditions, from dry concrete to water-containing media.

[0050] The 3D imaging and visualization terminal 6 serves as the output interface for the system's results. Its core task is to transform the 3D dielectric constant data generated by the signal processing unit 5 into an intuitive, interactive, and quantitatively valuable engineering diagnostic report. The 3D imaging and visualization terminal 6 first performs voxel rendering on the 3D dielectric constant volume data, then uses an advanced visualization engine to construct a true 3D structural model that can be arbitrarily rotated, scaled, and dynamically sliced. Users can use custom profile tools to "perspectively" observe the interior of the structure. The software also provides multiple rendering modes (such as isosurface rendering and volume drawing transparency adjustment) to highlight targets with different properties, such as reinforcing bars, voids, and water-bearing areas, thereby transforming abstract numerical data into easily understandable visual images and greatly improving the intuitiveness of the detection results.

[0051] Building upon precise visualization, the terminal integrates an intelligent post-processing analysis module, achieving a leap from "imaging" to "diagnosis." This module utilizes advanced image segmentation algorithms (such as region growing and level set methods) and feature extraction techniques to automatically identify and separate various abnormal structures in the 3D model. Subsequently, the system's built-in geometric metrology engine accurately calculates key parameters: it calculates the 3D central axis and diameter of the reinforcing bars through spatial topology analysis, determines the minimum distance from the surface of the reinforcing bars to the edge of the component (i.e., the protective layer thickness) through distance field analysis, directly outputs the volume of voids through region marking and voxel statistics, and accurately quantifies the width, length, and 3D spatial orientation of cracks through skeletonization and normal analysis. Finally, all this information—including the interactive 3D model, multi-angle slice views, and detailed parameter quantification tables—is automatically integrated to generate a well-structured and conclusive professional diagnostic report, providing direct and reliable data support for the safety assessment and maintenance decisions of engineering structures.

[0052] The working principle of this invention is based on the technical route of combining ultra-wideband electromagnetic wave scanning with three-dimensional full-wave inversion. Through the process of "data acquisition → inversion reconstruction → visualization diagnosis", it realizes non-destructive, quantitative, three-dimensional imaging of the internal structure of non-homogeneous media.

[0053] First, the system efficiently acquires massive amounts of high-quality raw echo data through a highly synchronized electronic scanning and data acquisition system. After system startup, the timing control and electronic scanning module 3 acts as the command center, executing a preset time-division multiplexing sequence with nanosecond-level precision to drive each transmitting antenna in the multi-transmitter module 1 to radiate ultra-wideband pulsed electromagnetic waves sequentially. After the electromagnetic waves penetrate the surface of the measured medium, they will be reflected, scattered, and attenuated when they encounter targets with different dielectric constants, such as steel bars, voids, and cracks. At this time, all receiving antennas in the multi-channel receiving module 2, under the global triggering of the data acquisition and synchronization system 4, simultaneously capture these reflected echo signals carrying internal information. This synchronous acquisition mechanism ensures the consistency of massive channel data in time and phase, laying a solid foundation for subsequent high-precision imaging. At the same time, the multi-band excitation strategy enables the system to simultaneously acquire complementary information that combines penetration depth (low frequency) and resolution capability (high frequency).

[0054] Then, the system utilizes GPU-accelerated core algorithms to process and invert massive amounts of data, directly converting the echo signal into a three-dimensional dielectric constant distribution model. Signal processing unit 5 receives the raw data and first performs multi-band signal preprocessing (including time-frequency analysis, interference suppression, and dispersion compensation) to improve the signal-to-noise ratio. Then, a multi-level fusion algorithm integrates the advantages of data from different frequency bands. Subsequently, the computational core enters the three-dimensional full-wave inversion process: this algorithm uses the entire waveform information (including amplitude, phase, and travel time) as constraints, and under the parallel acceleration of the GPU, it iteratively optimizes the dielectric constant model until the residual between the forward-modeled data and the measured data is minimized. This process directly solves the inverse problem of electromagnetic wave propagation, ultimately outputting a high-resolution, quantitative three-dimensional dielectric constant spatial distribution.

[0055] Ultimately, the system transforms the physical model into intuitive engineering diagnostic conclusions through a 3D visualization and quantitative analysis terminal. The 3D imaging and visualization terminal performs voxel rendering on the inverted 3D dielectric constant volume data, generating interactive 3D images that can be arbitrarily rotated and cut, making the internal structure readily apparent. Furthermore, the terminal integrates advanced image segmentation and feature extraction algorithms to automatically identify targets such as reinforcing bars, cracks, and voids, and accurately calculate their key geometric parameters. Finally, the system automatically generates a professional diagnostic report integrating the 3D model, sliced ​​views, and quantitative data, thus completing the entire process of providing accurate, efficient, and intuitive assessment of the health status of engineering structures.

[0056] This invention also introduces a non-contact electromagnetic detection method using the above system, comprising the following steps: 1) Arranging the multi-transmitter module 1 and the multi-channel receiver module 2 in a non-contact manner above the target under test; 2) Exciting each transmitting antenna sequentially according to the timing control and electronic scanning module 3, and synchronously acquiring the echo signals of all receiving antennas during each excitation; 3) Preprocessing the acquired multi-channel echo signals; 4) Inputting the preprocessed data into the signal processing unit 5, executing the GPU-accelerated three-dimensional full-wave inversion algorithm, and inverting to obtain the three-dimensional dielectric constant distribution model inside the target under test; 5) Visualizing and quantitatively analyzing the three-dimensional dielectric constant distribution model in the three-dimensional imaging and visualization terminal.

[0057] The above method includes a velocity calibration step before executing the three-dimensional full-wave inversion algorithm: the propagation velocity of electromagnetic waves in the measured medium is calibrated on-site by using a pre-set calibration target.

[0058] Example 1: During routine inspections of the bridge deck of an elevated bridge in a certain city, engineers used the non-contact electromagnetic detection system of this invention for precise assessment. First, the integrated probe was suspended 10 centimeters below the bridge deck, and electromagnetic wave velocity was calibrated on-site using pre-embedded standard test blocks. After system startup, the timing control module drove the 4×4 transmitting antenna array to intelligently switch between dual frequency bands in time-division multiplexing mode, working in conjunction with the 8×8 receiving antenna array to complete data acquisition over a 1-square-meter area within 30 seconds. The signal processing unit, relying on a GPU parallel computing architecture, first performed time-frequency analysis and adaptive filtering on the multi-band data. Then, through a regularized constrained three-dimensional full-wave inversion algorithm, it reconstructed a three-dimensional dielectric constant distribution model with centimeter-level accuracy within 3 minutes. The three-dimensional visualization terminal generated an interactive structural model using volume rendering technology, automatically identifying the three-dimensional spatial coordinates of the reinforcing bars, quantifying the protective layer thickness distribution, and accurately locating a 120cm³ concrete spalling area caused by reinforcing bar corrosion. Finally, a professional diagnostic report integrating the three-dimensional model, defect parameters, and safety level was output, providing complete data support for precise bridge maintenance.

[0059] Example 2: In the factory inspection of carbon fiber composite wing skin in aerospace manufacturing enterprises, this system is integrated into the end effector of a six-axis industrial robot for automated quality inspection. The system performs full-coverage scanning of the component surface at a constant 15 cm interval. For each measurement point, the deep medium response is obtained through a high-frequency excitation signal and a nanosecond-level synchronous acquisition system. In the data processing stage, a multi-band fusion strategy is adopted. The defect signal-to-noise ratio is enhanced through feature-level joint analysis and image-level wavelet fusion. The dielectric properties of each ply are accurately reconstructed based on a full-wave inversion algorithm using an improved alternating direction multiplier method. The detection system successfully identified a 15 cm² delamination defect located at a depth of 2.1 mm in the second ply and a 0.8 mm metal inclusion on the surface. The three-dimensional visualization platform accurately outputs the defect volume, burial depth, and boundary coordinates through adaptive threshold segmentation and spatial topology analysis, and generates a digital inspection and certification report that meets aerospace standards. This achieves accurate diagnosis of internal defects in composite materials from discovery and location to quantification.

[0060] Example 3: In urban old building renovation projects, it is necessary to accurately detect a 30-centimeter-thick reinforced concrete load-bearing wall to determine the distribution of embedded PVC conduits, metal water pipes, and reinforcing bars, providing a precise basis for subsequent trenching and wiring. On-site detection employs another hardware implementation of the system of this invention—using a high-precision vector network analyzer (VNA) as the core transceiver and data acquisition unit, replacing the dedicated pulse transmission and high-speed acquisition system in the aforementioned embodiments.

[0061] Before testing began, engineers connected the system's fixed multi-antenna array to the VNA's test port via an RF switching box, forming a complete bistatic radar-based testing system. The VNA, acting as the signal excitation source, transmitted a high-purity sinusoidal sweep signal in a stepped-frequency continuous wave mode within the 1.2GHz to 2.4GHz range, controlled by its internal phase-locked loop synchronization clock. This signal was distributed to the various transmitting antennas in the multi-transmitter module via an RF switch. When a transmitting antenna radiated electromagnetic waves towards the wall, all receiving antennas synchronously captured the signal reflected back through the internal structure and transmitted it back to the VNA's receiving port via a multi-channel RF switch. The VNA's built-in high dynamic range receiver performed amplitude and phase detection on the echo signals from each channel, outputting high-precision S-parameter data. This process, under the unified scheduling of the timing control module, traversed all transmitting antennas in a time-division multiplexing mode, ultimately obtaining a complete full-matrix scattering parameter dataset.

[0062] After data acquisition, the signal processing unit performs an inverse Fourier transform on the frequency domain data output by the VNA, converting it into an equivalent time-domain impulse response waveform. Background subtraction and multi-band fusion preprocessing are then performed. Based on this high-fidelity waveform data, a GPU-accelerated 3D full-wave inversion algorithm directly inverts the 3D distribution model of the dielectric constant inside the wall through iterative optimization. The 3D imaging and visualization terminal clearly presents the internal structure: three 20mm diameter PVC pipes run vertically, with the internal air cavity appearing as a significant low-dielectric-constant region; a 15mm diameter galvanized water pipe, due to its high conductivity, appears as a high-contrast target with clear boundaries in the dielectric constant model; the position and spacing of the double-layer steel mesh are also accurately reconstructed. The system automatically completes image segmentation and feature extraction, accurately outputting the 3D central axis coordinates, burial depth, and pipe diameter parameters of each pipeline, and generating a detailed pipeline distribution diagnostic report, providing complete and reliable information on underground concealed facilities for subsequent renovation and construction.

[0063] In summary, this invention innovatively proposes a non-contact electromagnetic detection system integrating a fixed multi-antenna array, electronic scanning technology, and a three-dimensional full-wave inversion algorithm. This system achieves synchronous transmission and reception of multi-band electromagnetic waves through a highly integrated hardware architecture. Combined with nanosecond-level precision timing control and signal synchronization mechanisms, it ensures excellent phase consistency and signal-to-noise ratio for massive amounts of acquired data. The core algorithm employs GPU-accelerated three-dimensional full-wave inversion technology, breaking through the limitations of traditional imaging methods. It can directly reconstruct the high-resolution three-dimensional distribution of dielectric constant within the medium. Finally, through an intelligent three-dimensional visualization and quantitative analysis platform, the physical parameters are transformed into intuitive structural models and precise geometric dimensions, enabling accurate diagnosis of internal defects in complex engineering structures from detection and location to quantification. This provides a revolutionary technical means for the field of non-destructive testing.

[0064] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Although the applicant has described the present invention in detail with reference to preferred embodiments, those skilled in the art should understand that any modifications or equivalent substitutions made to the technical solutions of the present invention cannot depart from the spirit and scope of the present invention and should be covered within the scope of the claims of the present invention.

Claims

1. A non-contact electromagnetic detection system based on the 3DFWI algorithm and the dielectric constant inversion algorithm, characterized in that, include: The multi-transmit module (1) consists of multiple transmitting antennas and is used to generate and radiate ultra-wideband pulsed electromagnetic waves. The multi-channel receiving module (2) consists of multiple receiving antennas and together with the multi-transmitting module (1) forms a fixed antenna array; the timing control and electronic scanning module (3) is configured with time-division multiplexing control logic to sequentially excite each transmitting antenna in the multi-transmitting module (1) and synchronously trigger the multi-channel receiving module (2) to perform data acquisition; the data acquisition and synchronization system (4) is used to synchronously acquire and record the echo signals received by the multi-channel receiving module (2); the signal processing unit (5) has a built-in hardware acceleration computing module; and the three-dimensional imaging and visualization terminal (6); wherein, the signal processing unit (5) is configured to: execute the three-dimensional full-wave inversion algorithm, process the acquired echo data, and invert the three-dimensional dielectric constant distribution model inside the target under test.

2. The non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm according to claim 1, characterized in that, The multi-transmit module (1) and the multi-channel receiver module (2) are integrated into a fixed transceiver probe.

3. The non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm according to claim 1, characterized in that, The timing control and electronic scanning module (3) achieves nanosecond-level switching of the transmitting antenna through a high-speed radio frequency switch array, and ensures that only one transmitting antenna in the multi-transmitting module (1) is excited at any given time, while all receiving antennas in the multi-channel receiving module (2) collect data in parallel.

4. The non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm according to claim 1, characterized in that, The multi-transmission module (1) is configured to switch the transmission signal between multiple preset center frequency points; the signal processing unit (5) is further configured to perform spectrum separation and weighted fusion processing on the data collected at different center frequencies.

5. The non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm according to claim 1, characterized in that, The hardware acceleration computing module is a GPU, and the three-dimensional full-wave inversion algorithm is accelerated using the parallel computing architecture of the GPU.

6. The non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm according to claim 5, characterized in that, The three-dimensional full-wave inversion algorithm is optimized using a multi-scale or multi-frequency iteration strategy to reduce its dependence on the initial model.

7. The non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm according to claim 1, characterized in that, The data acquisition and synchronization system uses a high-precision phase-locked loop or atomic clock as a synchronization source to ensure that the transmission and acquisition events are strictly synchronized in time.

8. The non-contact electromagnetic detection system based on the 3DFWI algorithm and dielectric constant inversion algorithm according to claim 1, characterized in that, The three-dimensional imaging and visualization terminal is configured to: display the inverted three-dimensional dielectric constant distribution model in voxel format, and automatically and quantitatively calculate the geometric parameters of internal defects through image segmentation and feature extraction algorithms.

9. A non-contact electromagnetic detection method using the system described in any one of claims 1-8, characterized in that, Includes the following steps: 1) The multi-transmitter module (1) and the multi-channel receiver module (2) are arranged above the target under test in a non-contact manner; 2) The timing control and electronic scanning module (3) excites each transmitting antenna in sequence according to the timing, and synchronously collects the echo signals of all receiving antennas during each excitation; 3) The collected multi-channel echo signals are preprocessed; 4) The preprocessed data is input to the signal processing unit (5) and the GPU-accelerated three-dimensional full-wave inversion algorithm is executed to invert the three-dimensional dielectric constant distribution model inside the target under test; 5) The three-dimensional dielectric constant distribution model is visualized, rendered and quantitatively analyzed in the three-dimensional imaging and visualization terminal.

10. The non-contact electromagnetic detection method according to claim 9, characterized in that, Before executing the three-dimensional full-wave inversion algorithm, a velocity calibration step is also included: the propagation velocity of electromagnetic waves in the measured medium is calibrated on-site using a pre-set calibration target.

Images