CN-121806007-B - Ground penetrating radar space-time registration and detection method for hidden diseases

CN121806007BCN 121806007 BCN121806007 BCN 121806007BCN-121806007-B

Abstract

The invention relates to the technical field of ground penetrating radar data processing, in particular to a space-time registration and detection method of a ground penetrating radar for hidden diseases. The method comprises the following steps of S1, establishing a continuous mapping relation between radar signals and a geographic space, S2, establishing a local track index mechanism based on anisotropic measurement, S3, establishing a space-time joint correction space based on error prior distribution, S4, parameter inversion based on maximum posterior probability estimation constraint, S5, establishing a registration quality assessment and signal characteristic decoupling mechanism, and S6, establishing a dielectric prior driven dual physical attribute index and a self-adaptive discrimination model. The invention overcomes the influence of GPS positioning error and track swing, realizes centimeter-level physical space alignment, obviously reduces false alarm caused by registration deviation, improves the signal-to-noise ratio under the background of complex cities by introducing physical constraint of frequency dimension, and obviously reduces false alarm rate.

Inventors

YAN YU
Zhang Daikang
DU YANLIANG
WU DIFEI
XU FEI
SUN LIJUN

Assignees

同济大学

Dates

Publication Date: 20260508
Application Date: 20260310

Claims (6)

1. A space-time registration and detection method of a ground penetrating radar for hidden diseases is characterized by comprising the following steps: S1, establishing a continuous mapping relation between radar signals and a geographic space; s2, constructing a local track index mechanism based on anisotropic measurement; s3, constructing a space-time joint correction space based on error prior distribution; s4, parameter inversion based on maximum posterior probability estimation constraint; S5, establishing a registration quality evaluation and signal characteristic decoupling mechanism; S6, constructing a dielectric priori driven dual physical attribute index and a self-adaptive discrimination model; the step S2 is specifically performed by, Establishing a local dynamic coordinate system along tangential direction and normal direction of a radar track, constructing space-time corridor constraint by utilizing vehicle kinematic boundary and physical width of an array radar, defining an anisotropic distance measurement function fusing longitudinal speed and transverse deviation weight in a feasible domain, searching an optimal matching neighborhood of a monitoring track relative to a reference track, and overcoming registration errors caused by non-uniform sampling and non-rigid deformation of the track; the specific implementation process of the local track index mechanism is as follows: Constructing a local coordinate system, namely selecting track points from the continuous space coordinate sequence of the reference ground penetrating radar output in the step S1 as reference points Selecting corresponding candidate points from the continuous space coordinate sequence of the monitoring ground penetrating radar as monitoring points With reference track points As origin, tangentially along the track is defined as longitudinal Defined as transverse in the normal direction Will monitor the point Mapping coordinates to the local coordinate system ; The space-time corridor constraint is to set a physical feasible domain, and only candidate points meeting the following double constraints are reserved: The lateral constraint is that Wherein In the event of a lateral deviation, For the physical width of the array radar, Is a positioning tolerance; Longitudinal constraint of Wherein In order for the longitudinal deviation to be the same, For the maximum travel speed of the vehicle, The time interval between two adjacent scans of the ground penetrating radar is used for limiting the longitudinal maximum drift of the vehicle; anisotropic neighbor search is to construct a high-dimensional spatial index for the reference point by adopting BallTree algorithm, and calculate the monitoring point in the space-time corridor constraint And datum point Is the anisotropic distance of (2) : Wherein the method comprises the steps of Normalized weighting factors of longitudinal and transverse directions respectively are selected The smallest point is used as the best matching point, and the lateral deviation corresponding to the neighborhood of the best matching point is extracted Initial lateral offset serving as spatial coarse registration of two-phase data of reference ground penetrating radar and monitoring ground penetrating radar To the subsequent step.
2. The method for space-time registration and detection of the ground penetrating radar for hidden diseases according to claim 1, wherein the step S1 is specifically, Acquiring reference ground penetrating radar data and monitoring ground penetrating radar data acquired by the same target area in different periods, carrying out plane projection processing on the original discrete track data, and carrying out continuous reconstruction on the projected track to establish a one-to-one correspondence between a radar scanning channel and geographic space coordinates; The reference ground penetrating radar data refer to radar echo patterns obtained by first census of a target pipeline area by using a vehicle-mounted array ground penetrating radar system at a historical moment and GPS track data synchronously collected by the radar echo patterns, the monitoring ground penetrating radar data refer to data obtained by retesting the same area by using the same or the same type of vehicle-mounted ground penetrating radar system at the current moment, and the radar echo patterns and the GPS track data both comprise radar waveform data and GPS positioning data; The plane projection is processed into longitude and latitude coordinates of the original GPS data Conversion to universal transverse-axis mercator projection coordinates To eliminate the earth curvature effect; The continuous reconstruction is based on radar track number Linear interpolation is carried out on projected coordinates as independent variables, and finally continuous space coordinate sequences corresponding to each path of radar data one by one are output A unified coordinate reference is provided for subsequent spatial registration.
3. The method for space-time registration and detection of the ground penetrating radar for hidden diseases according to claim 1 is characterized in that the step S3 is specifically to execute subspace alignment based on surface reflection phases aiming at random jitter acquired by radar signals; The method comprises the steps of extracting a first wave jump point of each channel of signal, namely a maximum amplitude phase point, based on subspace alignment of surface reflection phases, namely vertical phase alignment, aiming at radar antenna jitter caused by road surface fluctuation, uniformly correcting the surface reflection moments of all scanning channels to a time zero point by using a cross-correlation alignment algorithm, and eliminating vertical non-rigid deformation caused by acquisition system jitter; the probability feasible domain is based on the statistical property of GPS positioning errors And physical width of array radar Deriving confidence boundaries for parameter inversion, defining lateral offsets Is a search interval of (a) The method comprises the following steps: Wherein, the For the initial lateral offset calculated in the best matching point in step S2, As a result of the confidence coefficient, Is the standard deviation of GPS positioning error.
4. The method for space-time registration and detection of ground penetrating radar for hidden diseases according to claim 3, wherein step S4 is specifically, The GPS positioning error distribution characteristic is modeled as a prior probability regular term, the radar waveform similarity is modeled as a likelihood function, and a global objective function integrating physical prior and data observation is constructed; Prior modeling will be true lateral offset Considered as random variables, obeys Gaussian distribution Wherein For the coarse registration offset amount, The standard deviation of GPS positioning error; constructing an objective function, defining a fine-registered global cost function For a weighted sum of the relevance term and the regularization term: Wherein, the To monitor the lateral offset from the reference data Pearson correlation coefficients between the radar waveforms in the lower overlap region, As a physical penalty term based on GPS errors, Super parameters for adjusting the weight; The optimization algorithm is solved to solve the optimal physical offset in the feasible domain 。
5. The method for space-time registration and detection of ground penetrating radar for hidden diseases according to claim 1, wherein step S5 is specifically, The method comprises the steps of carrying out effective judgment on a registration result by utilizing a correlation evaluation index, extracting radar echo signals which finish spatial registration only when effective physical overlapping exists between reference data and monitoring data, carrying out signal transformation processing on the echo signals to realize characteristic decoupling, and obtaining instantaneous amplitude attribute representing reflected energy intensity and instantaneous frequency attribute representing medium attenuation characteristic; The registration quality evaluation and signal characteristic decoupling mechanism comprises the following specific implementation processes: the registration quality is evaluated as the optimal physical offset obtained according to the step S4 Performing lateral position translation correction on the monitoring ground penetrating radar data to finish the fine registration of physical space, setting a correlation threshold value If the global maximum correlation coefficient is calculated Judging that effective overlapping exists, and extracting the registered single-channel radar waveform in the overlapping area as a real signal ; When it is determined that there is significant overlap, for the real signal Resolving signals using Hilbert transform construction The corresponding expression is: Wherein the method comprises the steps of Representing the hilbert transform operator, In imaginary units, the instantaneous amplitude And instantaneous frequency The calculated expressions of (a) are respectively: 。
6. the method for space-time registration and detection of a hidden disease-oriented ground penetrating radar according to claim 5, wherein step S6 is specifically, Based on the dielectric response mechanism of the aqueous medium, respectively constructing an energy focusing index for representing the echo energy and a red shift index for representing the spectrum shift, utilizing the statistical distribution characteristics of the datum period data to establish a self-adaptive floating threshold value, constructing an anomaly discrimination model with double condition constraint, and realizing quantitative identification and space positioning of the hidden micro leakage signal; the specific implementation process of the index construction and discrimination model is as follows: Energy focus index : Wherein the method comprises the steps of For the energy at the present moment of time, Characterizing dielectric contrast induced energy mutations for background reference energy; Red shift index : Wherein the method comprises the steps of For the frequency of the current moment of time, The background reference frequency represents the amount of shift of the center frequency to low frequency relative to the background spectrum; Self-adaptive threshold setting, namely calculating the average value of index sequences of road sections without diseases in a reference period Standard deviation of The mean value of Comprising energy mean value And frequency mean The standard deviation Including the standard deviation of energy And standard deviation of frequency Setting dynamic threshold value, and energy enhancement threshold value as The frequency attenuation threshold is Wherein Confidence coefficients of energy and frequency respectively; The double criterion expression is: in the formula, The method is a binarization judgment result for concealing diseases, wherein 1 represents that abnormality exists, and 0 represents that no abnormality exists; and (3) with Respectively an energy focus index and a red shift index at the current moment, And (3) with The corresponding adaptive thresholds are respectively determined to be leakage anomaly only when the energy index is significantly higher than the upper limit and the red-shift index is significantly lower than the lower limit.

Description

Ground penetrating radar space-time registration and detection method for hidden diseases Technical Field The invention relates to the technical field of ground penetrating radar data processing, in particular to a space-time registration and detection method of a ground penetrating radar for hidden diseases. Background With the rapid advancement of the global urbanization process, urban underground spaces have become a complex network for carrying pipelines such as traffic, energy, communication, drainage and the like. However, this vast subsurface system is facing multiple challenges of infrastructure aging, geological environment migration, and high-strength ground loading, which induces the propagation and spread of various subsurface hidden diseases. The problems of road collapse, underground cavity, soil loosening, pipeline leakage and the like not only cause huge economic loss, but also form direct threat to public safety. Unlike visible diseases such as ground cracks or ruts, the underground hidden diseases tend to have extremely strong concealment in the early stage, and the development process is slow and is not easily perceived by the traditional surface inspection means until the underground hidden diseases develop into catastrophic ground collapse accidents. Therefore, how to realize early discovery, accurate positioning and dynamic monitoring of underground hidden diseases has become an issue of common concern in the fields of civil engineering and urban disaster prevention and reduction. Among the numerous nondestructive inspection techniques, ground penetrating radar is recognized as one of the most effective means for detecting road subsurface diseases by virtue of its excellent penetration ability into nonmetallic media, high-resolution shallow imaging level, and sensitivity to dielectric constant differences. In particular to a vehicle-mounted array ground penetrating radar system, through integrating a multi-channel antenna array, a high-precision positioning and attitude determining system and a high-speed data acquisition unit, full-coverage type fault scanning can be carried out on urban roads at normal running speed on the premise of not blocking traffic, and the detection efficiency and the data coverage rate are greatly improved. Although the technology of the vehicle-mounted ground penetrating radar is relatively mature, the technology still faces a serious technical bottleneck in the actual detection of specific diseases such as early tiny leakage of underground pipelines. Because of small water diffusion range and weak physical property difference caused by early leakage, the radar spectrum of single static scanning is difficult to distinguish from background clutter of surrounding underground complex media. Therefore, timing comparison analysis based on the same-region multi-period data becomes a key to identify such evolving diseases. However, this technical route is limited in engineering landing to the following problems: First, non-repeatability of the on-board acquisition trajectory results in spatiotemporal misalignment. Although the vehicle-mounted system is provided with a satellite positioning module, the vehicle-mounted system is limited by positioning errors of a civil navigation system and road traffic flow influence, a vehicle is difficult to strictly maintain a driving line consistent with a reference track during retest, and nonlinear transverse swing and offset often occur. This physical spatial misalignment causes the radar images acquired two times before and after to be not on the same cross section, resulting in direct image comparison analysis failure. Second, conventional registration and detection algorithms are poorly adapted. The existing ground penetrating radar image registration technology relies on significant feature points such as pipeline hyperbola peaks or horizon inflection points, but is difficult to be applied to even road sections with missing features. Meanwhile, a time domain amplitude difference method commonly used in the industry is extremely sensitive to registration residual errors, and small space dislocation can generate high-amplitude false artifacts at strong reflection interfaces such as road structure layers, so that real weak leakage signals are covered. In addition, the existing method often ignores the absorption characteristic of the water medium to the electromagnetic wave high-frequency component, namely the frequency red shift effect, and lacks a discrimination mechanism based on multidimensional physical properties. In summary, the prior art has obvious defects in the aspect of realizing accurate spatial alignment of multi-period vehicle-mounted radar data and weak water signal extraction. Therefore, there is a need for a space-time co-registration and detection method that can accommodate lateral vehicle track offset, has coarse-to-fine step alignment capability, and incorporates multi-dimensional physical attribute fea