Highway recessive disease three-dimensional integrated rapid diagnosis intelligent system

Abstract

The invention belongs to the technical field of intelligent traffic, and particularly relates to a highway recessive disease three-dimensional integrated rapid diagnosis intelligent system which comprises an ultra-wideband high-speed frequency hopping MIMO radar system platform used for realizing the balance between the detection depth and the resolution and obtaining the optimal resolution in the whole measurement depth range; a continuous and effective high-precision positioning technology platform for a complex environment and a navigation signal shielding rejection environment is used for realizing continuous and effective high-precision relative positioning of equipment, accurately corresponding acquired data to highway mileage piles and Beidou coordinates, and carrying out mutual association and fusion; a three-dimensional radar data depth information perception and identification technology platform is used for realizing automatic disease detection, disease type identification and disease grade evaluation. A recessive disease three-dimensional reconstruction technology platform is used for adaptively optimizing and adjusting reconstruction parameters for reconstruction of different types of diseases.

Description

Technical Field

[0001] The invention belongs to the field of intelligent transportation technology, and specifically relates to a three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases. Background Art

[0002] As one of the most important transportation facilities, highways are closely related to social development and people's lives. In recent years, road subsidence accidents have occurred frequently in domestic cities. However, the current "Highway Technical Condition Assessment Standard" (JTGH20-2007) only evaluates the road sign damage rate, and the evaluation of the roadbed technical condition is also based on the speculative judgment of the internal structure based on the apparent characteristics, without directly evaluating the internal structural defects of the road.

[0003] In fact, invisible diseases of the internal structure of the road do not necessarily directly reflect on the outside of the road structure. Compared with the apparent diseases of the road surface, the consequences of hidden diseases inside the highway are more serious, and of course, it is more difficult to detect them early. Ground penetrating radar (GPR) is a very efficient non-destructive testing method. In road engineering applications, the ground penetrating radar transmits high-frequency pulse electromagnetic waves through the transmitter antenna. The electromagnetic waves are transmitted into the road surface structure. When encountering medium interfaces with different dielectric properties and underground target surfaces, reflection, diffraction and refraction will occur. By analyzing the reflected echo signals received by the ground penetrating radar receiver, the parameters of the road structure layer, the location and degree of the disease can be analyzed.

[0004] However, there are still the following difficulties in detecting invisible road diseases through radar: the existing ultra-wideband high-speed frequency-hopping MIMO radar system is large in size and weight; the rapid diagnosis intelligent equipment cannot synchronously and quickly obtain accurate geographic location information over a large range in real time during operation, and the navigation continuity is poor in environments where navigation signals are blocked, such as tunnels, overpasses, forests, etc.; the high-definition data of the road surface and the data collected in parallel by multi-channel radars are very large, the efficiency of manual processing is low, and the environment is complex, making it "inaccurate"; the characteristics of three-dimensional radar data are quite different from those of stereo vision and point cloud data, and the existing three-dimensional reconstruction of radar data based on radar image characteristics and signal characteristics only uses its time domain information, while the frequency domain of the image data is not effectively utilized.

[0005] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgment or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases to solve the above-mentioned problem of radar detection of hidden highway diseases.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases, comprising:

[0009] Ultra-wideband high-speed frequency-hopping MIMO radar system platform, used to achieve a balance between detection depth and resolution, and obtain the best resolution over the entire measurement depth range;

[0010] The continuous and effective high-precision positioning technology platform in complex environments and navigation signal blocking and denial environments is used to achieve continuous and effective high-precision relative positioning of equipment, accurately correspond each collected data with highway mileage posts and Beidou coordinates, and correlate and fuse them with each other;

[0011] 3D radar data depth information perception and recognition technology platform, used to automatically detect diseases, identify disease types and assess disease levels;

[0012] The technology platform for three-dimensional reconstruction of hidden diseases is used to adaptively optimize and adjust reconstruction parameters for reconstruction of different types of diseases.

[0013] Preferably, the ultra-wideband high-speed frequency-hopping MIMO radar system platform takes the ultra-wideband high-speed frequency-hopping MIMO radar as the core, integrates RTK precise positioning module, photoelectric measuring wheel and high-definition camera, and forms a multi-information comprehensive acquisition platform including multi-channel MIMO antenna array module, parallel multi-channel wide-band transceiver front-end technology module, multi-channel high-speed data acquisition, transmission and storage integrated technology module.

[0014] Preferably, the multi-channel MIMO antenna array module realizes an ultra-wideband, low-profile and miniaturized directional antenna through multi-mode coupling technology and LCR circuit / absorbing material loading technology to support the system to achieve large depth, high resolution and dense scanning grid; through the antenna phase correction technology, a stable radiation pattern within the 200-2000MHz frequency band is guaranteed; through the array coupling leakage suppression technology, the array element coupling is prevented from affecting the system detection characteristics.

[0015] Preferably, the parallel multi-channel wide-band transceiver front-end technology module moves the small-bandwidth signal generated by the DDS circuit to different local oscillator frequencies through the DDS+DS cyclic up-conversion spread spectrum technology and the orthogonal discrete frequency coding sequence design method, and then realizes frequency expansion through the principle of frequency splicing; the entire transceiver front-end circuit includes four parts: a baseband circuit, a stepped frequency signal source part composed of a large-frequency local oscillator signal generation module and an orthogonal modulation circuit, an FPGA control system part, and a transmitter and receiver part.

[0016] Preferably, the main control machine of the multi-channel high-speed data acquisition, transmission and storage integrated technology module is based on FPGA+multi-channel ADC as the core, supplemented by automatic gain control circuit, synchronous clock generation circuit, communication interface and power distribution module; the task of FPGA is to generate the whole machine working timing signal, the control signal of MIMO antenna array configuration, the transmission channel gain setting signal, the control signal of multi-channel high-speed ADC module, the working start signal of the ranging wheel and RTK positioning terminal, the random sequence of orthogonal discrete frequency coded waveform, and the interface control logic of data acquisition, high-speed data sorting and storage control, data preprocessing and high-speed network port data transmission of the ranging wheel and RTK positioning terminal.

[0017] Preferably, the continuous, effective and high-precision positioning technology platform for complex environments and navigation signal obstruction and denial environments ensures accurate positioning of intelligent devices in complex electromagnetic environments by studying the adaptive anti-interference performance of INS / Beidou deep-coupled combined navigation terminals; a multi-dimensional combined navigation model is constructed by studying the joint ranging wheel and RTK / INS, and tracking and filtering are performed using multi-source information error state quantities to achieve Beidou deep-coupled combined navigation, thereby solving the problem of poor navigation continuity in navigation signal obstruction and denial environments such as tunnels, overpasses, and forests.

[0018] Preferably, the continuous effective high-precision positioning technology platform for complex environments and navigation signal blocking denial environments includes a BeiDou-3 new signal high-precision positioning and measurement module and a BeiDou-3 precise positioning terminal equipment research and development module; the BeiDou-3 new signal high-precision positioning and measurement module adopts the BeiDou-3 precise positioning algorithm to improve deformation tracking performance, and the specific steps of the algorithm are as follows:

[0019] Step 1, establish a carrier phase double difference equation group; linearize the nonlinearity of the equation group to obtain a linearized equation;

[0020] Step 2: The linearized equation is processed using the EKF prior estimation process, which is divided into three cases:

[0021] (1) If the current epoch is the first epoch, the prior estimated state matrix and its covariance matrix are updated as follows:

[0022] x k = [ x p , r y p , r z p , r N 1 N 2 ... N N ] T

[0023] ;

[0024] (2) If the current epoch is not the first epoch, and based on step 5, the timing is used to determine whether the periodic smoothing time has been reached; if not, the prior estimated state matrix and its covariance matrix are updated as follows:

[0025] x k − = [ x k − 1 , r y k − 1 , r z k − 1 , r N k − 1 , 1 N k − 1 , 2 ... N k − 1 , N ] T

[0026] ;

[0027] (3) If the current epoch is not the first epoch and the periodic smoothing time is reached, the prior estimated state matrix and its covariance matrix are updated as follows:

[0028] x k = [ x p , r y p , r z p , r N k − 1 , 1 N k − 1 , 2 ... N k − 1 , N ] T

[0029] ;

[0030] Step 3, perform the EKF status update process;

[0031] Step 4: Use LAMBDA / MLAMBDA to search for shaping ambiguities and obtain the three-dimensional position fixed solution of the current epoch;

[0032] Step 5. Save the solution results and judge the convergence of the time series formed by the solution results through the variance change method; if the current epoch is judged to be converged, output the convergence result; otherwise, enter the next epoch solution.

[0033] Preferably, the three-dimensional radar data deep information perception and recognition technology platform adopts the following technologies: three-dimensional radar data deep information perception technology, abnormal intelligent recognition technology based on physical model and data dual drive, and a new road disease rating method; the new road disease rating method provides a new basis for scoring the internal structure of the road, specifically including: subjective scoring basis for the internal structure of the road, establishing the relationship between the subjective score and the objective measurement value, calculating the weight coefficient of the invisible disease of the internal structure of the road, and rating the condition of the internal structure of the road.

[0034] Preferably, the three-dimensional reconstruction technology platform for hidden diseases uses the following steps to process the ground penetrating radar data of multiple parallel survey lines after conventional processing:

[0035] Step 1: Perform difference frequency processing on the target echo signal and the reference signal received by the radar to obtain the radar echo equation after the difference frequency; process the difference frequency signal to obtain the distance information between the target and the radar;

[0036] Step 2: Perform Fourier transform on the horizontal azimuth dimension data; perform Fourier transform on the vertical azimuth dimension data;

[0037] Step 3: Take the center of the imaging area as the reference point, take the phase of this point as the reference phase, then calculate the phase compensation amount by measuring the difference between the phase of other positions and the reference phase, and finally perform phase compensation on the image;

[0038] Step 4: interpolate the compensated data into a three-dimensional data volume;

[0039] Step 5: Perform three-dimensional inverse Fourier transform to obtain a preliminary reconstructed imaging area;

[0040] Step 6: Perform amplitude normalization processing on the acquired ground penetrating radar attribute volume;

[0041] Step 7: Use K-Means clustering algorithm to divide the data into two categories: background data and effective reflection data;

[0042] Step 8: Calculate the optimal amplitude value of the effective reflection data through adaptive optimization, which is the visualization amplitude threshold of the GPR attribute body after normalization;

[0043] Step 9: Combine the amplitude threshold and the normalized attribute volume, and use relevant software to achieve high-accuracy three-dimensional visualization of the hollow target.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases of the present invention aims to solve the problems of large size and weight of existing ultra-wideband high-speed frequency hopping MIMO radar systems. It proposes low-profile ultra-wideband antenna technology, orthogonal multi-channel ultra-wideband high-speed frequency hopping source technology, multi-channel parallel transceiver front-end technology and high-speed multi-channel data acquisition and storage technology. By developing and integrating functional modules such as low-profile ultra-wideband MIMO array antenna, RF transceiver front-end, real-time signal acquisition, high-speed and large-capacity data transmission, and GPU and CPU SOC architecture coordination, it develops an ultra-wideband high-speed frequency hopping MIMO radar system hardware platform for intelligent highway maintenance, achieves a better balance between detection depth and resolution, and obtains the best resolution over the entire measurement depth range.

[0046] (2) The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases of the present invention aims to solve the problems that the rapid diagnosis intelligent equipment cannot synchronously and quickly obtain a wide range of accurate geographic location information in real time during operation, and the navigation continuity is poor in tunnels, overpasses, forests and other navigation signal blocking and denial environments. The high-precision positioning technology of the photoelectric measuring wheel and INS / Beidou-3 dual-frequency RTK combination is proposed, and a low-cost embedded multi-frequency RTK terminal for Beidou-3 precise positioning is developed. The multi-dimensional combined navigation model is constructed by combining RTK positioning information and INS information such as high-precision ranging wheels. The error state quantity of multi-source information is used for tracking and filtering to achieve continuous and effective high-precision relative positioning of the equipment. The collected data are accurately matched with the highway mileage posts and Beidou coordinates, and are correlated and fused with each other to ensure accurate positioning in complex electromagnetic environments and navigation signal blocking and denial environments.

[0047] (3) The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases of the present invention addresses the problem that the high-definition road surface data and multi-channel radar parallel acquisition data are very large, the manual processing efficiency is low, and the environment is complex, and the detection is difficult. Therefore, an adaptive clutter suppression method, a low sidelobe suppression method, and a multi-feature driven deep learning neural network intelligent data analysis method are proposed to realize automatic detection of diseases, identification of disease types, and evaluation of disease levels, thereby providing maintenance support for highway maintenance plans.

[0048] (4) The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases of the present invention aims at the problem that the characteristics of three-dimensional radar data are quite different from those of stereo vision and point cloud data. The existing three-dimensional reconstruction of radar data based on radar image characteristics and signal characteristics only utilizes its time domain information, while the frequency domain of image data is not effectively utilized. A three-dimensional reconstruction method based on multiple characteristics of radar signals is provided to adaptively optimize and adjust reconstruction parameters for reconstruction of different types of diseases. The method first performs difference frequency processing on the target echo signal received by the radar and the reference signal to obtain the difference frequency echo, and then performs Fourier transform on it in the two-dimensional scanning direction, and then uses the imaging center point as the reference point to perform phase compensation on the data, and interpolates the compensated data, and then performs three-dimensional inverse Fourier transform on the interpolated data, and then extracts effective reflection data through clustering method, and then uses adaptive optimization to calculate the optimal amplitude value parameters of the effective reflection data, and finally achieves high-accuracy three-dimensional reconstruction based on the parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the modules of the rapid diagnosis intelligent system of the present invention;

[0050] Figure 2 It is the overall structure diagram of the ultra-wideband high-speed frequency-hopping MIMO radar system platform architecture;

[0051] Figure 3 This is a schematic diagram of the implementation principle of a large bandwidth stepped frequency signal;

[0052] Figure 4 It is a schematic diagram of the superheterodyne scheme;

[0053] Figure 5 It is the block diagram of the stepped frequency signal source system;

[0054] Figure 6 This is the block diagram of the BeiDou-3 dual-frequency RTK / INS deep-coupled integrated navigation and positioning technology;

[0055] Figure 7 It is a demonstration chart of the road internal structure damage condition score sheet. DETAILED DESCRIPTION

[0056] The following is a clear and complete description of the technical solution of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0057] Refer to the attached Figure 1 The present invention provides a three-dimensional integrated rapid diagnosis intelligent system for hidden road diseases, which mainly includes an ultra-wideband high-speed frequency-hopping MIMO radar system platform, a three-dimensional radar data depth information perception and recognition technology platform, a continuous effective high-precision positioning technology platform for complex environments and navigation signal blocking and rejection environments, and a three-dimensional reconstruction technology platform for hidden diseases based on digital twin technology. The specific introduction is as follows:

[0058] 1. An ultra-wideband high-speed frequency-hopping MIMO radar system platform

[0059] With ultra-wideband high-speed frequency-hopping MIMO radar as the core, it integrates RTK precise positioning module, photoelectric measuring wheel and high-definition camera to form a multi-information comprehensive acquisition platform, such as Figure 2 The radar system mainly includes a multi-channel MIMO antenna array module, a parallel multi-channel wide-band transceiver front-end technology module and a multi-channel high-speed radar host design. The details are as follows:

[0060] (1) Multi-channel MIMO antenna array module

[0061] To achieve fast scanning, the system adopts a frequency stepping system. The antenna units are composed of a 1×11-element receive / transmit array, and the transmit / receive antenna units are arranged alternately at intervals. MIMO technology is used to achieve antenna multiplexing, thus forming 21 channels. At the same time, in order to increase the scanning speed of the ground, the antenna array adopts a grid design. The radar is divided into two groups according to the sequential scanning of a pair of (transmit / receive) antenna arrays. The two groups of antennas scan simultaneously to speed up the progress. The antenna array multi-channel multiplexing technology ensures a spacing of 6 TX transmitting antennas to ensure that the two transmitting signals do not affect each other.

[0062] At the same time, through multi-mode coupling technology and LCR circuit / absorbing material loading technology, ultra-wideband, low-profile and miniaturized directional antennas are realized to support the system to achieve large depth, high resolution and dense scanning grids; through the antenna phase correction technology, a stable radiation pattern within the 200-2000MHz frequency band is guaranteed; through array coupling leakage suppression technology, the array element coupling is prevented from affecting the system detection characteristics.

[0063] The multi-channel MIMO antenna array module adopts the ultra-wideband, low-profile and miniaturized technology of the antenna unit and the ultra-wideband MIMO array antenna decoupling technology.

[0064] As mentioned above, the antenna unit has ultra-wideband, low profile and miniaturization technology. The wavelength of 200MHz reaches 1500mm, while the channel spacing is only 75mm. Therefore, the antenna system must first design a unit antenna with ultra-wide bandwidth, low profile and extremely small size. At the same time, the antenna also needs to obtain a stable radiation pattern within the range of 200MHz-2000MHz. To this end, the butterfly antenna is used as the basic unit and the following technologies are used to meet the requirements:

[0065] (a) Resistance loading technique

[0066] Through lumped or distributed resistance loading, the current reflection at the end of the antenna can be improved to a certain extent, thereby improving impedance matching and widening the effective working bandwidth of the antenna. In order to achieve the directional radiation characteristics of the antenna, a metal plate is loaded on the back of the dipole antenna. In order to reduce the impact of the loaded metal plate on the antenna matching, a metal back cavity is usually loaded at a quarter wavelength away from the antenna plane. The addition of the metal back cavity also introduces capacitive loading, which improves the impedance matching of the antenna to a certain extent. At the same time, resistance loading also helps to suppress the time domain signal tailing effect and improve the resolution of the antenna.

[0067] (b) Frequency selective surface

[0068] In order to ensure that the ultra-wideband antenna maintains good radiation performance in a wide frequency band, the frequency selective surface needs to have a wide-band characteristic and meet the following conditions: (1) low-frequency electromagnetic waves are not affected, that is, the antenna can transmit electromagnetic waves in the low-frequency part; (2) prevent the passage of high-frequency electromagnetic waves, that is, the antenna fully reflects electromagnetic waves in the high-frequency part. Therefore, the frequency selective surface is required to have a low-pass characteristic or a band-stop characteristic. First, the equivalent circuit analysis of the FSS is performed, and then the filtering characteristics of the square ring patch type frequency selective surface are determined.

[0069] The admittance of the square ring patch type frequency selective surface is:

[0070]

[0071] when hour, , at this time the impedance of the frequency selective surface is infinite:

[0072]

[0073] The corresponding frequencies are:

[0074]

[0075] It can be concluded from the formula that as long as L and C are changed, the position of the resonance point will be affected, and the filtering characteristics will be manifested as a band-stop at the frequency point. The simulation shows that the filtering characteristics of the square ring patch type frequency selective surface are less than -10dB at 2-3GHz, indicating that the transmission of electromagnetic waves is blocked in this working frequency band, and the filtering characteristics are played. At the same time, after the dipole antenna is loaded with the frequency selective surface, the phase of the electromagnetic wave at a specific frequency point can be changed, solving the problem of antenna radiation pattern distortion.

[0076] (c) Absorbing material technology

[0077] Based on the optimization of electromagnetic band gap and frequency selective surface, the present invention provides a calculation and simulation method of absorbing materials. The absorbing material technology can absorb the radiation from the back of the antenna and weaken the influence of phase compensation. Matlab-CST joint simulation is used to solve the problems of large number of CST simulation grids and long calculation time. The dielectric constant, magnetic permeability, dielectric loss tangent and magnetic loss tangent of the absorbing material are calculated by NRW method.

[0078] Electromagnetic simulation software will first fit the data, which will eventually increase the number of iterations in the simulation. In order to achieve fast simulation of CST, the above data is imported into Matlab to fit the data, and the fitted parameters are imported into CST with more parameters than the original data, thereby shortening the simulation time.

[0079] The ultra-wideband MIMO array antenna decoupling technology used in this system has the transmitting antenna and receiving antenna arranged in a row, dividing the mutual coupling effect between the transmitting antenna and the transmitting antenna and between the transmitting antenna and the receiving antenna.

[0080] (a) Isolation between transmitting antenna and transmitting antenna

[0081] Through the analysis of radiation characteristics, the coupling between antenna units is divided into two types: one is caused by space coupling electromagnetic waves, and the other is caused by floor surface waves. This technology mainly improves the isolation between the two transmission channels through two methods: suppressing the transmission of surface waves through electromagnetic band gaps; and suppressing the propagation of space waves through single negative metamaterials.

[0082] The single negative metamaterial is introduced as follows:

[0083] In nature, materials can be divided into four categories according to their constitutive parameters, namely, equivalent dielectric constant and equivalent magnetic permeability: (1) Equivalent dielectric constant and equivalent magnetic permeability All materials are positive; (2) and All materials have negative values; (3) For negative values Materials with positive values; (4) For positive value The material is negative. In (3) and (4), the electromagnetic wave will attenuate sharply in the material and show stopband characteristics. The simulation shows that the scattering parameter of the single negative material is less than -10dB in 2.9-4.7GHz, and the electromagnetic wave cannot be transmitted. After the single negative material is loaded between the antennas, the isolation between the antennas is significantly reduced, which effectively suppresses the transmission of space waves. Therefore, applying this material to the isolation between antennas can effectively suppress the influence of space waves.

[0084] (b) Isolation between the transmitting antenna and the receiving antenna

[0085] According to the principle of surface wave propagation, when a surface wave propagates to any discontinuous plane, it will produce traveling wave diffraction and radiate energy into free space. Therefore, a non-coplanar dielectric substrate and ground plane are used to suppress the propagation of surface waves, thereby improving the isolation between the transmitting antenna and the receiving antenna. The transmitting antenna and the receiving antenna are arranged in a V-shaped array to suppress the transmission of surface waves. The radiation areas of the transmitting antenna and the receiving antenna only partially intersect, which can effectively reduce the coupling between the transmitting antenna and the receiving antenna and improve the isolation between the transmitting branch and the receiving branch.

[0086] At the same time, by comparing the isolation between the transmitting antenna and the receiving antenna when they are collinear and when they are not (distance difference s), it can be seen that when the transmitting antenna and the receiving antenna are not collinear, when s is equal to l / 2 (l is the spacing between antenna units), the isolation is the largest. At the same time, in order to achieve the reuse of antennas, the cross distribution helps to make full use of the internal space of the ground penetrating radar, facilitate wiring, and improve the isolation between the transmitting and receiving antennas without significantly increasing the array size.

[0087] (2) Parallel multi-channel broadband transceiver front-end technology module

[0088] This technology moves the small bandwidth signal generated by the DDS circuit to different local oscillator frequencies, and then achieves frequency expansion through the principle of frequency splicing. While increasing the signal bandwidth, it retains the fast frequency switching characteristics of DDS and can meet the requirements of large bandwidth and fast frequency switching step frequency signal generation. Figure 3The figure shows a schematic diagram of the implementation principle. It can be seen from the figure that a large-bandwidth step frequency signal can be obtained by time-domain splicing of multiple narrow-band step frequency signals with the same bandwidth and uniformly changing center frequencies. Combined with the DDS plus mixer solution, the DDS circuit generates a small-step baseband frequency signal, and the local oscillator circuit generates a large-frequency local oscillator signal. By mixing a large-frequency local oscillator signal with a small-step baseband signal, the corresponding sub-band step frequency signal can be obtained. By changing the local oscillator in turn, the required large-bandwidth step frequency signal can be finally spliced. The entire transceiver front-end circuit consists of four parts: a step frequency signal source part composed of a baseband circuit, a local oscillator circuit and an orthogonal modulation circuit, an FPGA control system part, and a transmitter and receiver part. The specific introduction of each part is as follows:

[0089] ① Baseband circuit

[0090] Small step baseband signal generation module such as Figure 5 As shown in the upper left. The baseband circuit mainly includes DDS chip, reference source, synchronization source, amplifier, low-pass filter and other parts, and its control signal is provided by the FPGA in the system control circuit. The reference source AD9510 generates various working clocks for other devices by dividing or multiplying the external input clock signal; the synchronization source AD9510 uses the synchronization output signal of I-channel DDS chip as the main synchronization signal to generate 3-channel slave synchronization signals and provide them to FPGA and 2 DDS chips respectively to ensure that the design meets the phase coherence; the two DDS chips AD9910 generate two differential signals respectively under the corresponding working clock, synchronization signal and control signal, and then convert them into single-ended signals by amplifier AD8001 and output them to the low-pass filter, and finally obtain the required -100MHz~100MHz I / Q small step baseband frequency signal.

[0091] ② Step frequency signal source

[0092] The step frequency signal source is mainly composed of a local oscillator circuit and an orthogonal modulation circuit. The overall system block diagram is shown in Figure 5 shown.

[0093] (a) Local oscillator circuit

[0094] The local oscillator circuit, i.e., the high-frequency local oscillator signal generation module, is mainly composed of a reference frequency source, a clock distribution chip, a PLL circuit, a variable gain amplifier chain, a high-speed switch, etc. The reference frequency source provides a reference frequency signal of 100MHz, which is sent to the PLL circuit to generate a combined signal after reasonable distribution, and the output amplitude of each point frequency is adjusted through the variable gain amplifier. Finally, under the control of the local oscillator selection signal, a high-speed switch selects a certain output as a high-frequency local oscillator signal.

[0095] (b) Quadrature modulation circuit

[0096] It is mainly composed of a broadband amplifier, a low-pass filter and an orthogonal modulator. The broadband operational amplifier and the low-pass filter together form a signal conditioning circuit, which is responsible for accurately adjusting the DC and AC levels of the I and Q signals input to the orthogonal modulator so that the two have good amplitude and phase consistency, thereby improving the quality of the output signal of the orthogonal modulator. The orthogonal modulator divides the local oscillator signal by 90° and mixes it with the I and Q small step frequency signals that are orthogonal to each other, and then synthesizes the resulting signals, so that the required sideband signals are superimposed on each other, while the useless sidebands cancel each other out.

[0097] ③ FPGA control system

[0098] FPGA controls the DDS chip to generate a small step frequency signal, which enters the orthogonal modulation circuit after filtering and amplification. At the same time, FPGA controls the high-speed switch in the local oscillator circuit to select the corresponding large-frequency local oscillator signal and send it to the modulation circuit for mixing and time-domain splicing with the small step frequency signal to generate a 190MHz~1990MHz broadband step signal, which is sent to the transmitting antenna after filtering, power adjustment and up-conversion to 200MHz~2GHz, where the intermediate frequency signal is fixed at 10MHz. The echo signal is filtered and amplified, and then down-converted by the mixer and filtered to generate an intermediate frequency signal. The orthogonal demodulator sends the collected intermediate frequency signal data back to the FPGA control part for data processing.

[0099] ④ Transmitter and receiver part

[0100] (a) Key technologies of receiver

[0101] In the stepped radar system, the common stepped receiver mainly includes the following two structures: superheterodyne receiver and zero intermediate frequency receiver. However, due to the inherent flicker noise, local oscillator leakage, DC deviation and other problems of zero intermediate frequency receiver, it is not as stable as the superheterodyne receiver. Therefore, superheterodyne receivers are mostly used in mobile communication systems. The basic superheterodyne structure includes the RF part and the intermediate frequency part.

[0102] After the signal is received by the receiving antenna, it enters the LNA through the limiter. After the signal is amplified, it is filtered by the mirror filter and then enters the mixer for mixing. After the intermediate frequency signal is obtained, it is filtered and amplified, and the original IQ two-way baseband signal is demodulated in the baseband circuit part.

[0103] (b) Transmitter key technologies

[0104] The superheterodyne scheme is also often used in broadband transmitters. Compared with the direct up-conversion scheme, the superheterodyne scheme has a better isolation. The output signal of the power amplifier is well isolated from the local oscillator signal, and the frequencies are different, thus avoiding the occurrence of local oscillator pulling. Usually, the output frequency of the modulator in the superheterodyne scheme is low, so it can achieve good modulation accuracy. The scheme diagram is shown in the attached figure. Figure 4 As shown, the orthogonal modulation part is implemented in the frequency source part.

[0105] (3) Multi-channel high-speed radar host, that is, a multi-channel high-speed data acquisition, transmission and storage integrated technology module

[0106] The main control machine is composed of FPGA + multi-channel ADC as the core, supplemented by automatic gain control circuit, synchronous clock generation circuit, communication interface and power distribution module, etc. Among them, the task of FPGA is to generate the whole machine working timing signal, the control signal of MIMO antenna array configuration, the transmission channel gain setting signal, the control signal of multi-channel high-speed ADC module, the working start signal of the ranging wheel and RTK positioning terminal, the random sequence of orthogonal discrete frequency coded waveform, and the data acquisition of the ranging wheel and RTK positioning terminal, high-speed data sorting and storage control, data preprocessing and high-speed network port data transmission interface control logic, etc.

[0107] The multi-channel high-speed data acquisition, transmission and storage integrated technology module adopts the control logic of MIMO antenna array configuration and multi-channel ADC acquisition control logic; among them,

[0108] The control logic of the MIMO antenna array configuration is as follows: the configuration of the 11-transmit 11-receive MIMO antenna array is reconfigurable, and the transmit and receive antennas are synthesized into a linear array with a measurement point spacing of 7.5 cm to meet the requirements of lateral dense measurement lines. In order to avoid mutual interference between transmitted signals, the transmit antenna array is divided into two groups (1-5) and (6-11). The antenna spacing between the two groups is 5 in one transmission cycle, and each group adopts a dual-transmit and dual-receive mode using the aforementioned orthogonal discrete frequency coding waveform. The entire control logic is generated by the FPGA to control the RF switch of the antenna array. In addition, the synthesized 21-channel array antenna structure triggers an acquisition when the measuring wheel detects that it has traveled 2 cm. The host computer integrates and processes the received data to obtain a three-dimensional C-Scan image.

[0109] The specific logic of multi-channel ADC acquisition control is as follows: the main control machine uses multiple high-speed ADC chips LTC2285 to synchronously acquire 6 channels of echo signals with a center frequency of 10MHz and an intermediate frequency signal bandwidth of 1MHz. The control parameters of data acquisition are written to the intermediate frequency receiver system by the host computer through the network port to ensure that the sampling trigger, sampling delay, sampling length and the frequency range of the step frequency signal source are all controllable. In fact, after the system is powered on, the ADC begins to continuously convert the intermediate frequency signal, but the read and write signals generated by the internal logic of the FPGA can control when the external FIFO stores the quantized data and how long the data is stored, thereby indirectly achieving the control of the delay and sampling length. When designing the system, in addition to configuring 4 hardware FIFOs outside the FPGA, a 1K*64 internal FIFO should be generated inside the FPGA using the IP core provided by the company to form a secondary cache structure to better solve the problem of easy loss due to the high frequency of the AD converter.

[0110] 2. Continuous, effective and high-precision positioning technology platform in complex environments and navigation signal blocking and denial environments

[0111] BeiDou-3 / GNSS technology is used to achieve precise positioning and high-precision measurement, but this method has a slow data update rate, the radar antenna needs to maintain a relatively uniform speed during data collection, and spatial position information is easily lost. Therefore, this technology platform uses a joint dual-frequency RTK and high-precision ranging wheels to achieve high-precision positioning, specifically including the BeiDou-3 new signal high-precision positioning and measurement module and the BeiDou-3 precise positioning terminal equipment research and development module.

[0112] (1) BeiDou-3 new signal high-precision positioning and measurement module

[0113] The new high-precision positioning and measurement method of BeiDou-3 signal is adopted to analyze the matching relationship between BeiDou-3 signal structure, signal propagation and signal reception processing, and proposes signal processing technologies such as average correlation and carrier error compensation signal capture, genetic particle filter multipath suppression code loop, etc., to solve the problems of BeiDou signal multi-peak, weak signal capture and poor tracking ability, which can improve the signal capture speed and peak detection probability. The new high-precision positioning and measurement method of BeiDou-3 signal includes multiple algorithms and technologies, as follows:

[0114] ① BeiDou-3 BOC(n,n) unambiguous acquisition and tracking algorithm

[0115] The orthogonal demodulation method is used to generate two carriers with a right spatial phase angle to achieve the purpose of removing the carrier of the signal after down-conversion; then, the currently generated subcarrier is multiplied by the pseudo-random code, and the spectrum is expanded to generate a local BOC signal. Taking each PRN code chip width as a reference, it is divided into two parts, and the first half plus the second half are zeroed to form an odd branch signal. Similarly, the second half plus the first half are zeroed to form an even branch signal; the odd branch signal is coherent with the received signal, and then the coherence value is lagged by half a code chip and the inverse is taken. The new correlation function is multiplied with the original value, modulo, and then added to the value without modulo, so as to achieve the purpose of peak elimination.

[0116] The tracking ambiguity of BOC is eliminated by reconstructing the self-correlation function, mainly by improving the code tracking loop. Therefore, the receiver still uses the traditional tracking loop to complete carrier synchronization. The received signal first strips the I and Q local carriers generated by the receiver tracking loop. At the same time, the code loop locally generates SPRN1 and SPRNN as two local auxiliary codes and advances and lags them. After the received signal is correlated with the local code, the local code phase is adjusted through the phase detector and code NCO, thereby completing the unambiguous tracking of BOC.

[0117] ② BeiDou-3 dual-frequency RTK / INS deep coupling high-precision positioning and attitude determination technology

[0118] A multi-source deep coupling structure that integrates scalar deep combination and vector deep combination is adopted, that is, the channel is kept closed when the channel signal is good, and the channel is opened when the signal is severely weakened or interrupted, so as to realize adaptive disconnection and closure of the channel. The SNR-FFT method for estimating signal energy is proposed, which can quickly and accurately judge the signal attenuation and recovery, adaptively control the disconnection and recovery of the channel tracking loop, realize continuous tracking of Doppler and pseudo code when some signals are intermittent, and fast recovery when all signals are intermittent. The BeiDou-3 dual-frequency RTK / INS deep coupling combined navigation and positioning method is shown in the attached figure. Figure 6 shown.

[0119] The multi-antenna BeiDou-3 attitude measurement technology uses three BeiDou-3 satellite signal receiving antennas fixed on the surface of the carrier according to certain rules. On the premise that the three antennas can receive BeiDou-3 satellite signals, the satellite signals received by each BeiDou-3 antenna are processed to obtain carrier phase observation values ​​and code phase measurement values. The obtained measurement values ​​are combined with the least squares attitude solution algorithm to calculate the coordinates of the baseline vector between antennas in the local horizontal coordinate system, and finally the attitude of the carrier is solved through coordinate transformation.

[0120] ③ BeiDou-3 monitoring data precise time synchronization technology

[0121] It provides a precise time synchronization technology between BeiDou-3 base stations in a complex environment. The dual-chip synchronization interface technology is implemented through two device interfaces, one GPIO interface and one baseband event input (BB_EVENT) interface in the region. This interface is a physical link layer interface for raw observations or time synchronization.

[0122] The main chip monitors the auxiliary chip interface to send data frames. After the auxiliary chip receives the antenna signal normally and starts, the main chip sends a signal through the GPIO port and connects to the EVENT input interface of the main chip and the auxiliary chip at the same time. The EVENT interface of the main and auxiliary chips samples the input signal at the same time and generates an interrupt. The internal logic and firmware cooperate to achieve time synchronization of the input original observation quantity. When there is an external event (BB_EVENT), BB_EVENT is triggered to latch all RTC counters for software access. When the EVENT interface is implemented, the high-speed system clock is used to synchronize the external input signal. The synchronized signal generates a BB_EVENT status signal according to the software configuration; the BB_EVENT status signal is used as the external latch trigger source of the RTC module, and then determines whether to generate a BB_EVNET interrupt based on the interrupt mask setting.

[0123] The original observation synchronization technology is based on the dual-chip synchronization interface technology and realizes time synchronization through software calculation. After the auxiliary chip (L1 / B1) is running, L2 / B2 outputs a pulse through the GPIO pin, and the pulse signal is connected to the EVENT pin of the main and auxiliary chips. After the main and auxiliary chips receive the pulse signal on the EVENT pin, the hardware automatically latches the RTC value. After entering the EVENT interrupt service program, the software latches the RTC and write address again, and then the software calculates the time difference based on the two latched RTC values. The latched write address minus the calculated time difference can know the write address difference of the RF write TE FIFO in the two chips.

[0124] ④ Narrowband anti-interference and multipath suppression

[0125] (a) To remove narrowband interference, a frequency domain filtering method based on generalized extension interpolation is used;

[0126] The adaptive nulling of the narrowband anti-interference algorithm adopts the power inversion criterion and the iteration adopts the NLMS algorithm. The algorithm has two processes in implementation: one is the filtering process and the other is the weight adaptive updating process. The calculation process is divided into three steps: first, the initial weight is set, the input vector is filtered, and the filter output is obtained. Then, the error is calculated using the reference signal and the filter output. Finally, the weight is adjusted according to the error and the input vector.

[0127] (b) Multipath suppression technology, which uses genetic algorithms to optimize the resampling process of particle filters and combines them with Kalman filters to deal with signal amplitude issues, thereby effectively suppressing the multipath effect;

[0128] Particle filtering (PF) is a recursive approximation filtering theory based on random sampling methods. It uses state samples in the spatial model, i.e. particles, to represent the distribution and simulate the state probability density function to obtain the state minimum variance estimate. In order to solve the problems of PF degradation and particle fading in traditional resampling, the genetic algorithm is embedded in the resampling process of the particle filter. By simulating the evolutionary phenomenon in biology, the particle filter resampling method is optimized, overcoming the degradation problem of the particle filter and the particle diversity fading problem in resampling. At the same time, the linear problem of signal amplitude is processed by Kalman filtering (KF).

[0129] ⑤ BeiDou-3 precise positioning algorithm

[0130] Based on the research of dynamic and static algorithms, the state prior estimation process of the adaptive fusion dynamic and static algorithm EKF is adjusted based on the time period to improve the deformation tracking performance while meeting the accuracy. The specific steps of the dynamic and static adaptive fusion algorithm are as follows:

[0131] Step 1: Establish the carrier phase double difference equations. According to the nonlinearity of the equations, linearize them to get the equations;

[0132] Step 2: The linearized equation is processed using the EKF prior estimation process, which is divided into three cases:

[0133] 1) If the current epoch is the first epoch, the prior estimated state matrix and its covariance matrix are updated as follows:

[0134] x k = [ x p , r y p , r z p , r N 1 N 2 ... N N ] T

[0135]

[0136] 2) If the current epoch is not the first epoch, and based on step 5, determine whether the periodic smoothing time has been reached according to the timing; if not, the prior estimated state matrix and its covariance matrix are updated as follows:

[0137] x k − = [ x k − 1 , r y k − 1 , r z k − 1 , r N k − 1 , 1 N k − 1 , 2 ... N k − 1 , N ] T

[0138]

[0139] 3) If the current epoch is not the first epoch, if the periodic smoothing time is reached, the prior estimated state matrix and its covariance matrix are updated as follows:

[0140] x k = [ x p , r y p , r z p , r N k − 1 , 1 N k − 1 , 2 ... N k − 1 , N ] T

[0141]

[0142] Step 3: Perform the state update process of EKF (the posterior estimation process of the state matrix);

[0143] Step 4: Use LAMBDA / MLAMBDA to search for shaping ambiguities and obtain the three-dimensional position fixed solution of the current epoch;

[0144] Step 5: Save the solution results and judge the convergence of the time series formed by the solution results through the variance change method; if the current epoch is judged to be converged, output the convergence result; otherwise, enter the next epoch solution.

[0145] (2) BeiDou-3 precise positioning terminal equipment R&D module, including BeiDou-3 high-precision base station board design module, BeiDou-3 / GNSS low-power and low-cost precise positioning terminal design module and distance measuring wheel and RTK-GPS joint positioning module. The details are as follows:

[0146] ① BeiDou-3 high-precision reference station board design module

[0147] It mainly includes the following parts: power management module, clock module, baseband processing module, AD / DA module, and MAX2771 module.

[0148] The power module is responsible for power supply and circuit protection;

[0149] The clock module supports higher precision clock input;

[0150] The baseband processor module is mainly composed of baseband processing chip FPGA+DSP and data storage chips SDRAM, NVSRAM, and FLASH; FPGA expands UART, ADC, and DAC interfaces through IO ports, and is mainly responsible for high-speed signal acquisition and processing and data transmission; DSP bus is equipped with FLASH, SDRAM chips, etc., to complete data interaction with FPGA, and is responsible for the processing and data storage of high-speed signals collected by FPGA;

[0151] The AD / DA module is equipped with 6 channels of digital-to-analog conversion transmission channels and 6 high-speed analog-to-digital acquisition channels;

[0152] The MAX2771 module is used as a receiver module, and can capture and track after processing the output intermediate frequency data.

[0153] ② Beidou-3 / GNSS low-power and low-cost precision positioning terminal design module

[0154] The BeiDou-3 / GNSS low-power, low-cost precision positioning terminal is deployed on mobile monitoring targets such as ground-penetrating radar equipment to monitor displacement, angle change and other changes in ground-penetrating radar equipment. It is a terminal device that uses BeiDou-3, GPS, and GLONASS multi-system satellite positioning to accurately calculate displacement and angle changes. The terminal is based on the BeiDou-3 / GNSS multi-mode chip and uses innovative precision positioning and real-time deformation measurement technology to achieve centimeter-level dynamic monitoring accuracy. The mobile monitoring station is connected to the precision positioning base station to obtain high-precision reference position data. At the same time, it supports carrier phase output, adopts weak signal reception technology based on noise coupling theory in a tightly coupled environment, and uses the integrated tightly coupled optimization design of the antenna and receiving module to make the terminal OTA TIS close to the navigation chip conduction sensitivity.

[0155] ③ Joint positioning module of ranging wheel and RTK-GPS

[0156] The specific idea of ​​the joint positioning method using the ranging wheel and RTK-GPS is as follows: when the ground penetrating radar antenna collects data along the survey line, the high-precision ranging wheel continuously triggers the ground penetrating radar host unit during its movement, and controls the transmitting antenna to emit electromagnetic waves, while triggering the mobile station GPS host and recording the GPS time at this moment; the base station GPS and mobile station GPS data are post-differentially processed, and the track data on the ground penetrating radar image and the accuracy of the ranging wheel are combined. Based on the principle of time synchronization, the track data on the ground penetrating radar image and the GPS spatial position information are accurately matched, so that the total number of track data of the radar image has accurate position information, thereby realizing the precise positioning of the ground penetrating radar image, and the time synchronization accuracy is tested based on the consistency between the total number of tracks of the ground penetrating radar image and the number of GPS pulse events.

[0157] 3. 3D radar data depth information perception and recognition technology platform. This platform uses: 3D radar data depth information perception technology, abnormal intelligent recognition technology based on physical model and data dual drive, and a new road disease rating method. The details are as follows:

[0158] (1) 3D radar data depth information perception technology

[0159] Since the frequency range of waveform synthesis is 200MHz~2000MHz, up to 10 octaves. In such a large bandwidth, the antenna's directional patterns at low and high frequencies are quite different, which makes it difficult to achieve correction in the RF part. Therefore, it is necessary to perform systematic correction in the data domain for each baseband signal collected, including antenna dispersion compensation, intermediate frequency coupling suppression, and channel amplitude / phase correction.

[0160] Use a vector network analyzer to measure the group delay of waveform transmission between any two antennas, and then compensate for antenna dispersion.

[0161] In addition, the stepped frequency ground penetrating radar waveform belongs to the continuous wave system, which has serious signal leakage problems. The existence of coupled waves reduces the sensitivity of the receiver, causing saturation or even damage to the front end of the receiver. In order to reduce the influence of coupled waves and pulse compression on the echo signal of deep weak targets, the receiver performs gain control on the echo signal before sampling, suppresses the direct wave and amplifies the deep target echo to increase the dynamic range of the receiver.

[0162] The main factors affecting the channel amplitude and phase errors are antenna array elements, feed lines, analog devices at the receiving front end, and ADC devices. Ideally, the amplitude and phase of each channel are required to be consistent, but it is difficult to achieve complete consistency in the actual radar system characteristics. This technology first directly uses the original waveform to calculate the cross-correlation to obtain the delay difference, and uses the frequency domain to calculate the correlation function to improve the efficiency of the delay difference calculation, that is:

[0163] R ( m ) = FFT [ X 1 ( k ) X 2 ∗ ( k ) | X 1 ( k ) | | X 2 ( k ) | ]

[0164] Then the waveform delay caused by the linear phase-frequency error is estimated based on the known information. The waveforms of all channels are aligned to the 0 position by translation, so the estimated value of the high-order phase-frequency error satisfies:

[0165]

[0166] The corrected data is the I / Q data of each frequency point of the stepped frequency signal, and waveform compression is required to obtain the amplitude signal and delay information of the echo. Because the one-dimensional time domain waveform is the linear superposition of different delays of the time domain waveform corresponding to the spectrum envelope, the out-of-band of the synthetic spectrum is considered to be zero, and truncation discontinuities appear at both ends of the spectrum, similar to the "Gibbs" and "frequency domain leakage" phenomena, and the time domain waveform produces side lobes. In the SVA algorithm, spectrum extrapolation can compensate for the signal loss caused by weighting and maintain the resolution unchanged. The spectrum extrapolation super-resolution technology is used to suppress the range sidelobes. After obtaining the extrapolated spectrum, combined with the linear weighting method, a good sidelobe suppression effect can be obtained.

[0167] (2) Intelligent abnormality identification technology based on physical model and data dual drive

[0168] The present invention provides a defect detection and identification method based on physical mechanism and data dual drive, and proposes an improved deep learning model, namely Fast Regional Convolutional Neural Network (Fast R-ConvNets), in combination with the echo characteristics of ground penetrating radar. The main advantage of this model is that it can detect pavement defects with high accuracy and high recall without human assistance. In addition, the performance of this model in different pavement structures and materials is stable. Compared with the existing fast R-CNN model, the model proposed in the present invention can be regarded as its modification. The modification is divided into two parts: hyperparameter optimization to adapt to the detection task and algorithm improvement to improve detection accuracy. The first part is the optimization of anchor ratio and size, the purpose of which is to make the bounding box suitable for detecting pavement defects using ground penetrating radar data; the second part mainly includes the use of random pooling and layer sequence unit variance initialization, both of which have been verified to improve algorithm accuracy, but are not used in the existing fast R-CNN.

[0169] The fast R-ConvNet model proposed in the present invention extracts multiple two-dimensional images from a three-dimensional image to reflect its spatial features, that is, it does not directly process three-dimensional data, and has low classification and recognition capabilities, especially in the case of similar characteristics on B-Scan. In fact, the three-dimensional convolutional neural network (3D-CNN) has a stronger spatial-temporal feature learning ability of 3D data than the 2D CNN, learns different classification features from the 3D data body, and improves the classification recognition rate. On the basis of the original improved model, the present invention replaces the original 2D-CNN network with 3D-CNN, and replaces the input data B-Scan with C-Scan data. In addition, it should be noted that the size of the three-dimensional data body is different, so the cropped three-dimensional cubes of different sizes should be processed into the same size for importing into 3D-CNN. In the length and depth directions, the size of the cube can be simply changed by interpolation without affecting feature extraction, while in the width direction, the interpolation method may destroy the spatiotemporal features between parallel B-Scans. Therefore, it is necessary to use multiple mirror encoding (MME) to rearrange the three-dimensional cube data in the width direction to achieve equal-size three-dimensional data input.

[0170] (3) Providing a new road damage rating method

[0171] The formula uses the pavement condition index (PCI) as the evaluation index of pavement integrity and the pavement disease comprehensive damage rate (DR) as the survey index. The calculation formula is:

[0172]

[0173]

[0174] Where DR is the comprehensive damage rate of pavement diseases (%), is the weight of the i-th type of road damage, is the damaged area of ​​the i-th type of pavement disease, and A is the total damaged area of ​​asphalt pavement diseases.

[0175] The current Highway Technical Condition Assessment Standard (JTGH20-2007) only evaluates the road sign damage rate, but does not directly evaluate the road internal structure damage. Therefore, the present invention proposes a new scoring basis for the road internal structure:

[0176] ① Subjective scoring basis for road internal structure

[0177] In order to obtain objective and reasonable scoring rules, the expert survey method is first used to evaluate the damage rate of the internal structure of the road. In this way, the final score can be more reasonably evaluated and the maintenance countermeasures can be accurately obtained, and the scores of similar degree of damage can be approximate.

[0178] ② Establish the relationship between subjective scores and objective measurements

[0179] The collected hidden defects of the internal structure of the road are classified into cracks, looseness, high water content, and voids. The single deduction value is calculated according to the density and degree of each type of defect. The single deduction values ​​are then added up and subtracted to calculate the SDR as objective data. Figure 7 Establish the relationship between the subjective and the objective. When scoring, the hidden diseases of the internal structure of the road are all of a single degree of damage, a single density, and a single type. The average of the scores of the team members is calculated to determine the deduction values ​​for each type of damage, degree, and different damage densities.

[0180] ③ Calculation of weight coefficient of invisible disease of road internal structure

[0181] The weights of various hidden diseases in the road structure are determined by mathematical statistics. The evaluation and expert scoring results of several sections with similar disease types and densities are selected from the radar image database of the measured road, and the single deduction value SDRi is calculated. The weight coefficient is set as wi. The ISCI of several sections with hidden diseases in a road structure is calculated using the obtained weight value and single deduction value. If the deviation between the obtained ISCI and the average expert score is within an acceptable range, the obtained weight can be used as the weight value of the hidden disease of this type of road structure under the deduction ratio; otherwise, the single deduction value is adjusted and the weight is re-solved until it is acceptable. Then, sections with different distribution ratios of hidden diseases of this type of road structure are selected, and the above process is repeated to obtain the single deduction value of the hidden disease of this type of road structure and the weight function under different severity distribution ratios. For other types of hidden defects in the internal structure of roads, the above process can be repeated to obtain the single deduction points for different types of hidden defects in the road structure and the weight functions under different deduction ratios of severity.

[0182] ④ Rating of road internal structure condition

[0183] According to the obtained weight function, the final road internal structure condition index ISCI can be obtained. First, the modified road internal structure invisible disease damage rate DSR is calculated, and then the road structure internal condition index ISCI is fitted according to the damage condition rating. The technical condition level is determined according to the score range and divided into four types: excellent, good, medium and poor. The maintenance countermeasures suggestions of the expert group members for the hidden diseases of the road internal structure and the maintenance countermeasures currently used based on the road surface disease damage rate are integrated to improve the road structure technical condition evaluation system and obtain more accurate maintenance countermeasures.

[0184] 4. Hidden disease 3D reconstruction technology platform

[0185] This platform is based on a three-dimensional reconstruction method based on multiple features of radar signals, and adaptively optimizes and adjusts reconstruction parameters for reconstruction of different types of diseases. This method first performs frequency difference processing on the target echo signal received by the radar and the reference signal to obtain the echo after frequency difference, and then performs Fourier transform on it in the two-dimensional scanning direction. Then, the imaging center point is used as the reference point to perform phase compensation on the data, and the compensated data is interpolated. Then, a three-dimensional inverse Fourier transform is performed on the interpolated data, and effective reflection data is extracted through clustering methods. Then, the optimal amplitude value parameters of the effective reflection data are calculated using adaptive optimization, and finally, highly accurate three-dimensional reconstruction is achieved based on this parameter.

[0186] Specifically, for multiple parallel line ground penetrating radar data after conventional processing, the three-dimensional reconstruction method processing steps are as follows:

[0187] ①Difference frequency processing: Perform difference frequency processing on the target echo signal received by the radar and the reference signal to obtain the radar echo equation after the difference frequency processing. By processing the difference frequency signal, the distance information between the target and the radar is obtained.

[0188] ② Obtain frequency domain information: perform Fourier transform on the horizontal azimuth dimension data; perform Fourier transform on the vertical azimuth dimension data.

[0189] ③ Phase compensation: Use the center of the imaging area as a reference point for phase compensation. The basic idea of ​​phase compensation technology is to select a reference point in the imaging area, use the phase of this point as the reference phase, and then calculate the phase compensation amount by measuring the difference between the phase of other positions and the reference phase. Finally, perform phase compensation on the image to eliminate distortion and blur.

[0190] ④Synthesize three-dimensional data volume: interpolate the compensated data into a three-dimensional data volume.

[0191] ⑤ Preliminary three-dimensional imaging: Perform three-dimensional inverse Fourier transform to obtain the preliminary reconstructed imaging area.

[0192] ⑥ Normalization processing: To facilitate analysis, the acquired ground penetrating radar attribute volume is subjected to amplitude normalization processing.

[0193] ⑦ Cluster analysis: Use K-Means clustering algorithm to divide the data into two categories: background data and effective reflection data.

[0194] ⑧ Calculate the amplitude threshold parameter: Calculate the optimal amplitude value of the effective reflection data through adaptive optimization. This value is the visualization amplitude threshold of the GPR attribute body after normalization.

[0195] ⑨ Three-dimensional reconstruction: Combine the amplitude threshold and the normalized attribute volume, and use relevant software to achieve high-accuracy three-dimensional visualization of the hollow target.

[0196] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases, characterized in that: include: Ultra-wideband high-speed frequency-hopping MIMO radar system platform, used to achieve a balance between detection depth and resolution, and obtain the best resolution over the entire measurement depth range; The continuous and effective high-precision positioning technology platform in complex environments and navigation signal blocking and denial environments is used to achieve continuous and effective high-precision relative positioning of equipment, accurately correspond each collected data with highway mileage posts and Beidou coordinates, and correlate and fuse them with each other; 3D radar data depth information perception and recognition technology platform, used to automatically detect diseases, identify disease types and assess disease levels; The technology platform for three-dimensional reconstruction of hidden diseases is used to adaptively optimize and adjust reconstruction parameters for reconstruction of different types of diseases.

2. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 1 is characterized in that: The ultra-wideband high-speed frequency-hopping MIMO radar system platform is based on the ultra-wideband high-speed frequency-hopping MIMO radar, and integrates the RTK precise positioning module, the photoelectric measuring wheel and the high-definition camera to form a multi-information comprehensive acquisition platform including a multi-channel MIMO antenna array module, a parallel multi-channel wide-band transceiver front-end technology module, and a multi-channel high-speed data acquisition, transmission and storage integrated technology module.

3. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 2 is characterized in that: The multi-channel MIMO antenna array module realizes ultra-wideband, low-profile and miniaturized directional antenna through multi-mode coupling technology and LCR circuit / absorbing material loading technology to support the system to achieve large depth, high resolution and dense scanning grid; through the antenna phase correction technology, a stable radiation pattern within the 200-2000MHz frequency band is guaranteed; through the array coupling leakage suppression technology, the array element coupling is prevented from affecting the system detection characteristics.

4. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 2 is characterized in that: The parallel multi-channel wide-band transceiver front-end technology module uses the DDS+DS cyclic up-conversion spread spectrum technology and orthogonal discrete frequency coding sequence design method to move the small bandwidth signal generated by the DDS circuit to different local oscillator frequencies, and then realizes frequency expansion through the principle of frequency splicing; the entire transceiver front-end circuit includes four parts: a baseband circuit, a stepped frequency signal source part composed of a large frequency local oscillator signal generation module and an orthogonal modulation circuit, an FPGA control system part, and a transmitter and receiver part.

5. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 2 is characterized in that: The main control machine of the multi-channel high-speed data acquisition, transmission and storage integrated technology module is based on FPGA+multi-channel ADC as the core, supplemented by automatic gain control circuit, synchronous clock generation circuit, communication interface and power distribution module; the task of FPGA is to generate the whole machine working timing signal, the control signal of MIMO antenna array configuration, the transmission channel gain setting signal, the control signal of multi-channel high-speed ADC module, the working start signal of the ranging wheel and RTK positioning terminal, the random sequence of orthogonal discrete frequency coded waveform, and the interface control logic of data acquisition, high-speed data sorting and storage control, data preprocessing and high-speed network port data transmission of the ranging wheel and RTK positioning terminal.

6. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 1 is characterized in that: The continuous, effective and high-precision positioning technology platform for complex environments and environments with navigation signal obstruction and denial ensures accurate positioning of intelligent devices in complex electromagnetic environments by studying the adaptive anti-interference performance of the INS / Beidou deep-coupled combined navigation terminal. A multi-dimensional combined navigation model is constructed by studying the joint ranging wheel and RTK / INS, and tracking filtering is performed using multi-source information error state quantities to achieve Beidou deep-coupled combined navigation, thereby solving the problem of poor navigation continuity in environments with navigation signal obstruction and denial, such as tunnels, overpasses, bridges, and forests.

7. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 6 is characterized in that: The continuous effective high-precision positioning technology platform for complex environments and navigation signal blocking denial environments includes a BeiDou-3 new signal high-precision positioning and measurement module and a BeiDou-3 precise positioning terminal equipment research and development module; the BeiDou-3 new signal high-precision positioning and measurement module adopts the BeiDou-3 precise positioning algorithm to improve deformation tracking performance, and the specific steps of the algorithm are as follows: Step 1, establish a carrier phase double difference equation group; linearize the nonlinearity of the equation group to obtain a linearized equation; Step 2: The linearized equation is processed using the EKF prior estimation process, which is divided into three cases: (1) If the current epoch is the first epoch, the prior estimated state matrix and its covariance matrix are updated as follows: ； (2) If the current epoch is not the first epoch, and based on step 5, the timing is used to determine whether the periodic smoothing time has been reached; if not, the prior estimated state matrix and its covariance matrix are updated as follows: ； (3) If the current epoch is not the first epoch and the periodic smoothing time is reached, the prior estimated state matrix and its covariance matrix are updated as follows: ； Step 3, perform the EKF status update process; Step 4: Use LAMBDA / MLAMBDA to search for shaping ambiguities and obtain the three-dimensional position fixed solution of the current epoch; Step 5. Save the solution results and judge the convergence of the time series formed by the solution results through the variance change method; if the current epoch is judged to be converged, output the convergence result; otherwise, enter the next epoch solution.

8. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 1 is characterized in that: The three-dimensional radar data deep information perception and recognition technology platform adopts: three-dimensional radar data deep information perception technology, abnormal intelligent recognition technology based on physical model and data dual drive and a new road disease rating method; the new road disease rating method provides a new basis for scoring the internal structure of the road, specifically including: subjective scoring basis for the internal structure of the road, establishing the relationship between the subjective score and the objective measurement value, calculating the weight coefficient of the invisible disease of the internal structure of the road, and rating the condition of the internal structure of the road.

9. The three-dimensional integrated rapid diagnosis intelligent system for hidden highway diseases according to claim 1 is characterized in that: The three-dimensional reconstruction technology platform for hidden diseases uses the following steps to process the ground penetrating radar data of multiple parallel survey lines after conventional processing: Step 1: Perform difference frequency processing on the target echo signal and the reference signal received by the radar to obtain the radar echo equation after the difference frequency; process the difference frequency signal to obtain the distance information between the target and the radar; Step 2: Perform Fourier transform on the horizontal azimuth dimension data; perform Fourier transform on the vertical azimuth dimension data; Step 3: Take the center of the imaging area as the reference point, take the phase of this point as the reference phase, then calculate the phase compensation amount by measuring the difference between the phase of other positions and the reference phase, and finally perform phase compensation on the image; Step 4: interpolate the compensated data into a three-dimensional data volume; Step 5: Perform three-dimensional inverse Fourier transform to obtain a preliminary reconstructed imaging area; Step 6: Perform amplitude normalization processing on the acquired ground penetrating radar attribute volume; Step 7: Use K-Means clustering algorithm to divide the data into two categories: background data and effective reflection data; Step 8: Calculate the optimal amplitude value of the effective reflection data through adaptive optimization, which is the visualization amplitude threshold of the GPR attribute body after normalization; Step 9: Combine the amplitude threshold and the normalized attribute volume, and use relevant software to achieve high-accuracy three-dimensional visualization of the hollow target.

Images