CN114998645A - Road cavity form classification method based on three-dimensional GPR forward modeling technology - Google Patents

Road cavity form classification method based on three-dimensional GPR forward modeling technology Download PDF

Info

Publication number
CN114998645A
CN114998645A CN202210503794.8A CN202210503794A CN114998645A CN 114998645 A CN114998645 A CN 114998645A CN 202210503794 A CN202210503794 A CN 202210503794A CN 114998645 A CN114998645 A CN 114998645A
Authority
CN
China
Prior art keywords
dimensional
cavity
road
gpr
scan
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210503794.8A
Other languages
Chinese (zh)
Inventor
侯斐斐
王一军
张航
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Priority to CN202210503794.8A priority Critical patent/CN114998645A/en
Publication of CN114998645A publication Critical patent/CN114998645A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/10—Image acquisition
    • G06V10/16—Image acquisition using multiple overlapping images; Image stitching
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
    • G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Geometry (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Hardware Design (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Graphics (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

The invention provides a road cavity form classification method based on a three-dimensional GPR forward modeling technology, which utilizes GprMax3D software based on a three-dimensional time domain finite difference method principle to carry out three-dimensional GPR forward modeling simulation work of cavities with different forms aiming at typical road disease bodies represented by the cavities so as to obtain corresponding three-dimensional radar data bodies; the method has the advantages that the information extraction is carried out on the image characteristics of the data volume in different directions, and the road cavity form classification method based on the three-dimensional GPR forward modeling is jointly characterized, so that the spatial resolution and the measurement precision are higher; and then extracting SIFT characteristics and accurately classifying the forms of the cavity diseases in the road structure by adopting an SVM classifier. The method has important guiding significance for interpreting the three-dimensional GPR actual measurement image and improving the reliability of detection result judgment, and can provide guidance for analyzing actual measurement data of the urban road underground cavity radar.

Description

Road cavity form classification method based on three-dimensional GPR forward modeling technology
Technical Field
The invention relates to the technical field of ground penetrating radar image processing, in particular to a road cavity form classification method based on a three-dimensional GPR forward technology.
Background
With the rapid development of economy in China, road construction as an economic development artery is also rapidly developed. However, under the combined action of traffic load and environmental factors, permanent deformation and internal damage are accumulated continuously, so that multiple road surface collapse events occur frequently, and a warning clock is sounded for road safety problems. Road collapse is usually caused by gradual expansion of underground cavities, has concealment, outburst and harmfulness, and becomes a key point and a difficult point of urban management. The traditional road monitoring means hardly meets the requirements of large-scale and harmless cavity form evaluation and detection, and an Impulse type Ground Penetrating Radar (GPR) is used as a non-contact non-destructive system, has strong penetrating power, has the characteristics of real time, multiple dimensions, high precision and the like, and in view of various advantages, GPR becomes a preferred method in the field of underground cavity detection, and provides data support and technical basis for monitoring and governing road diseases. In the process of detecting the underground cavity, the method is vital to shape recognition of the cavity, the shape scale and the geometric size of the cavity reflect the collapse evolution speed change degree and the severity degree, and the state of the cavity in a three-dimensional (3D) space can be accurately sensed. Therefore, how to accurately and rapidly classify and identify the hole form is an important subject for road maintenance monitoring.
There are still three challenges: firstly, no relevant research is available for analyzing and identifying the three-dimensional morphological characteristics of road cavities; secondly, regarding the selection of the shape of the hollow hole, the existing research only discusses the hollow hole as a simple shape such as a rectangle or a circle, and does not relate to more complicated shapes, even irregular hollow hole targets; third, in the three-dimensional GPR research on road cavities, the existing work is mostly to model road cavities and analyze forward results of cavities based on the GprMax technology, or to distinguish the cavity target from other types of underground objects by feature comparison, and does not involve related understanding and discussion on the cavity morphology. Therefore, the road cavity form identification technology based on the three-dimensional GPR forward technology has good application prospect and practical value.
Disclosure of Invention
In order to solve the technical problems, the invention provides a method for interpreting and classifying road cavity forms based on a three-dimensional GPR forward modeling technology, which is used for acquiring a large amount of GPR typical cavity disease three-dimensional data, extracting cavity information and form characteristics, and realizing automatic identification and classification of cavities with different forms by adopting an SIFT feature descriptor and an SVM classifier.
In order to achieve the purpose, the invention adopts the technical scheme that:
the road cavity form classification method based on the three-dimensional GPR forward modeling technology is characterized by comprising the following steps of: the method comprises the following steps:
s1: three-dimensional GPR road hole modeling and three-dimensional GPR data acquisition,
constructing a three-dimensional road void simulation model by adopting GprMax3D, and acquiring 3DGPR void data;
s2: extracting images in different directions of the cavity form,
and acquiring a B-scan image set and a C-scan image set along xoy and xoz planes respectively;
s3: the combined characterization of the shape of the hollow cavity,
arranging the B-scan images and the C-scan images in sequence, and combining and splicing the B-scan images and the C-scan images into a new two-dimensional image;
s4: the SIFT characteristics are extracted, and the method comprises the following steps of,
extracting scale invariant feature transform features (SIFT features) of the new combined image;
s5: the training test of the SVM classifier is performed,
carrying out K-means clustering and bag-of-word model conversion on the extracted SIFT features in sequence, then obtaining a new image vector, and using the new image vector as the input of an SVM classifier to realize training and testing of the SVM classifier;
s6: the result of the classification of the shape of the cavity,
and after training and testing of the SVM classifier, outputting a classification result of the cavity form.
As a preferred technical scheme of the invention: in step S1: the three-dimensional road cavity simulation model comprises five layers, namely a first air layer, a second asphalt layer, a third concrete layer, a fourth sand gravel layer and a fifth road base layer.
As a preferred technical scheme of the invention: the roadbed simulated random medium establishes a soil layer model, and the soil dispersion material is defined as follows: the soil has a sand content of 50% and a clay content of 50%, and the sand density is 2.66g/cm 3 Clay bulk density of 2g/cm 3 And a volume water content ranging from 0.001 to 0.25, the above materials being randomly distributed in a volume of 2X 1.2X 1m 3 On the model of (2).
As a preferred technical scheme of the invention: the cavity forms include spherical cavity, rectangular cavity, cylindrical cavity and irregular cavity, spherical cavity, rectangular cavity and cylindrical cavity have smooth surfaces, irregular cavity sets up to uneven surface.
As a preferred technical scheme of the invention: the three-dimensional GPR system parameters comprise a spatial resolution parameter of 0.01m, a time window parameter of 14ns, an initial transmitting antenna coordinate parameter of (0.45m, 1.0m and 0.0m), an initial receiving antenna coordinate parameter of (0.35m, 1.0m and 0.0m), an antenna stepping distance parameter of (0.01m, 0m and 0m), a measuring point number parameter of 100, an excitation signal type of Ricker and an excitation signal frequency parameter of 800 MHz.
As a preferred technical scheme of the invention: in step S2: extracting two-dimensional data of a two-dimensional section diagram B-scan and a horizontal section diagram C-scan from a three-dimensional structure S of GPR hole data, and performing shape recognition of the underground hole by using combined data of the two-dimensional section diagram B-scan and the horizontal section diagram C-scan:
1) two-dimensional data are sequentially extracted from S along a plane parallel to xoz, and a plurality of section views S are obtained by imaging 1 ＝{B 1 ,B 2 ,B 3 ,…,B n }，n＝l y Δ y, wherein B n Represents eachZhang B-scan, S 1 Then a set of n B-scans, l y The method comprises the steps that the length of a y-axis of a model space is taken, delta y is a space step length in the y-axis direction, in addition, after B-scan is extracted, preprocessing operations such as direct wave removal, data standardization, depth and speed change correction and the like are carried out on each image, signals and background noise are removed, the B-scan is extracted at equal intervals, and display is carried out in a stacking mode;
2) two-dimensional data is extracted from S along a plane parallel to xoy, and a plurality of horizontal sectional images S are obtained by imaging 2 ＝{C 1 ,C 2 ,C 3 ,…,C m }，m＝l z ,/Δ z, wherein C m Represents each C-scan, S 2 Is a collection of m horizontal slices, l z To model the spatial z-axis length, Δ z is the spatial step in the z-axis direction.
As a preferred technical scheme of the invention: in step S3: fusing a plurality of B-scans and C-scans images, and screening 8B-scans { B ] from each cavity model 1 ,B 2 ,B 3 ,B 4 ,B 5 ,B 6 ,B 7 ,B 8 And 12C-scans { C } 1 ,C 2 ,C 3 ,C 4 ,C 5 ,C 6 ,C 7 ,C 8 ,C 9 ,C 10 ,C 11 ,C 12 And sequentially arranging and combining to form a new two-dimensional image, wherein the image set I is expressed as a formula (1), each three-dimensional hole is finally converted into an image with fused information characteristics,
Figure BDA0003636474600000031
as a preferred technical scheme of the invention: in step S4: searching key points by using an SIFT algorithm and calculating the direction of the key points, wherein the dimensionality of a feature vector of each key point is 128, the first 120 key points are taken, the matrix of a descriptor is 120 × 128, all the key points are clustered by using a K-means algorithm, the clustering number is 10, each picture is represented as a bag-of-words model, each picture can be uniformly represented by a vector with the size of 10 × 1, 123 hole images with different forms are obtained in total, the ratio of a training set to a testing set is 2:1, 82 training sets and 41 testing sets are obtained.
As a preferred technical scheme of the invention: in step S5: and sending the training set and the labels thereof into an SVM to train and test a supervised learning classifier.
As a preferred technical scheme of the invention: and classifying the test set by using the classifier trained in the step S5, and outputting a classification result.
Compared with the prior art, the invention has the beneficial effects that:
the method has the beneficial effects that the research of the patent has important reference value in the aspects of extracting the three-dimensional GPR cavity characteristics and realizing the interpretation of the cavity form. After the problem is solved, the related problem about the detection of the cavity diseases in the road structure in the real scene can be easily solved, which is the foundation for implementing and optimizing the compaction of the subsequent large road condition detection system.
Drawings
FIG. 1 is a schematic flow diagram of the present invention.
FIG. 2 shows GPR system parameters in the context of a road structure in accordance with the present invention.
Fig. 3 is a dielectric characteristic of the road structure of the present invention.
FIG. 4 is a diagram of a road structure simulation model according to the present invention.
FIG. 5 is a simulation model of holes in different embodiments of the present invention.
FIG. 6 is a two-dimensional cross-sectional view stack of spherical voids.
FIG. 7 is a horizontal cross-sectional view stack of spherical voids.
FIG. 8 is a two-dimensional cross-sectional view stack of rectangular voids.
FIG. 9 is a horizontal cross-sectional view stack of rectangular voids.
FIG. 10 is a two-dimensional cross-sectional view stack of cylindrical voids.
FIG. 11 is a horizontal cross-sectional view stack of cylindrical voids.
FIG. 12 is a two-dimensional cross-sectional view stack of irregular voids.
FIG. 13 is a horizontal cut-away view stack of irregular voids.
FIG. 14 is a graph showing the result of feature fusion for spherical voids.
Fig. 15 is a graph showing the result of feature fusion for rectangular holes.
FIG. 16 is a graph showing the result of feature fusion for cylindrical voids.
Fig. 17 is a graph showing the result of feature fusion of irregular voids.
FIG. 18 is a graph showing the classification result of the hole pattern.
List of reference numerals:
1. an air layer; 2. an asphalt layer; 3. a concrete layer; 4. a layer of sand gravel; 5. and (4) a road bed layer.
Detailed Description
The invention is described in further detail below with reference to the following detailed description and accompanying drawings:
road collapse is a common phenomenon of urban roads in recent years, threatening road safety and operation, and underground cavities are one of the main causes for road collapse. In order to clarify radar detection characteristic images of different forms of cavity diseases and accurately analyze radar detection data, GprMax3D software based on a three-dimensional time domain finite difference method principle is utilized, three-dimensional GPR forward modeling work of different forms of cavities is carried out aiming at typical road disease bodies represented by the cavities, corresponding three-dimensional radar data bodies are obtained, and information extraction and combined characterization are carried out on image characteristics of the data bodies in different directions; and then extracting SIFT characteristics and accurately classifying the forms of the cavity diseases in the road structure by adopting an SVM classifier. The method has important guiding significance for interpreting the three-dimensional GPR actual measurement image and improving the reliability of detection result judgment, and can provide guidance for analyzing actual measurement data of the urban road underground cavity radar.
The specific technical scheme is as follows:
the road cavity form classification method based on the three-dimensional GPR forward modeling technology comprises the following steps:
s1: three-dimensional GPR road hole modeling and three-dimensional GPR data acquisition,
adopting GprMax3D to build a three-dimensional road cavity simulation model and acquiring 3DGPR cavity data;
s2: extracting images in different directions of the cavity form,
and acquiring a B-scan image set and a C-scan image set along xoy and xoz planes respectively;
s3: the combined characterization of the shape of the hollow cavity,
arranging the B-scan images and the C-scan images in sequence, and combining and splicing the B-scan images and the C-scan images into a new two-dimensional image;
s4: the SIFT characteristics are extracted, and the method comprises the following steps of,
extracting scale invariant feature transform features (SIFT features) of the new combined image;
s5: the training test of the SVM classifier is performed,
carrying out K-means clustering and bag-of-word model conversion on the extracted SIFT features in sequence, then obtaining a new image vector, and using the new image vector as the input of an SVM classifier to realize the training and testing of the SVM classifier;
s6: the result of the classification of the shape of the cavity,
and after training and testing of the SVM classifier, outputting a classification result of the cavity form.
The method adopts a GPR three-dimensional forward modeling technique GprMax3D to be used for three-dimensional modeling simulation of a road structure, and researches underground cavity models in different forms so as to obtain three-dimensional data containing rich information.
The GPR system parameter settings are shown in fig. 2. The asphalt concrete layered road structure is shown in fig. 3, and comprises a first air layer 1, a second asphalt layer 2, a third concrete layer 3, a fourth gravel layer 4 and a fifth road base layer 5. Wherein the air layer 1 is 0.2m, the asphalt layer 2 is 0.1m, the concrete layer 3 is 0.2m, the gravel layer 4 is 0.1m, the roadbed layer is 0.6m, namely the soil layer is 0.6m, a random medium is simulated to establish a more vivid soil model about the fifth road base layer 5, and a series of soil dispersion materials are defined as follows: the soil has a sand content of 50% and a clay content of 50%, and the sand density is 2.66g/cm 3 The clay bulk density was 2g/cm 3 The volume water content is in the range of 0.001-0.25. Distributing the above materials in a volume of 2 × 1.2 × 1m 3 On the model (where the soil layers are randomly distributed). The whole road structure simulation model is shown in fig. 4.
The invention designs four common typical hollow forms which are arranged in a road structure model, including a spherical hollow, a rectangular hollow, a cylindrical hollow and an irregular hollow, as shown in fig. 5. The first three cavities have smooth surfaces, and the cavity on the noble side is set to be an uneven surface, so that the cavity is closer to the real underground cavity. To further show the void volume inside the model, the left column of fig. 5 shows the internal structure of the model at x ═ 1m, and the right column of fig. 5 shows the void display under random media.
The method does not directly analyze the three-dimensional GPR data, but respectively extracts a two-dimensional section B-scan and a horizontal section C-scan from the three-dimensional data, and uses the combined data of the two-dimensional section B-scan and the horizontal section C-scan to identify the form of the underground cavity, as shown in FIGS. 6 to 13. Extracting the two forms of two-dimensional data from the three-dimensional structure S of the GPR hole data:
1) two-dimensional data are sequentially extracted from S along the plane parallel to xoz, and a plurality of section images S are obtained by imaging 1 ＝{B 1 ,B 2 ,B 3 ,…,B n }，n＝l y Δ y, wherein B n Represents each B-scan, S 1 Then a set of n B-scans, l y To model space y-axis length, Δ y is the space step in the y-axis direction. In addition, after the B-scan is extracted, preprocessing operations such as direct wave removal, data standardization, depth and speed change correction and the like are carried out on each image, and signals and background noise are removed. The B-scan is extracted equally spaced and displayed in a stacked manner, as shown in fig. 3-6;
2) two-dimensional data is extracted from S along a plane parallel to xoy, and a plurality of horizontal sectional images S are obtained by imaging 2 ＝{C 1 ,C 2 ,C 3 ,…,C m }，m＝l z ,/Δ z, wherein C m Represents each C-scan, S 2 Is a set of m horizontal slices, l z To model the spatial z-axis length, Δ z is the spatial step in the z-axis direction. The C-scan stack is shown in fig. 3-6.
Fusing a plurality of B-scans and C-scans images, and screening 8B-scans { B ] from each cavity model 1 ,B 2 ,B 3 ,B 4 ,B 5 ,B 6 ,B 7 ,B 8 And 12C-scans { C } 1 ,C 2 ,C 3 ,C 4 ,C 5 ,C 6 ,C 7 ,C 8 ,C 9 ,C 10 ,C 11 ,C 12 And (4) sequentially arranging and combining the images to form a new two-dimensional image, wherein the image set I is expressed as a formula (1). Each three-dimensional cavity is finally converted into an image after information characteristic fusion, and fig. 14-17 show characteristic fusion results of four cavities with different forms.
Figure BDA0003636474600000061
Searching key points by using an SIFT algorithm and calculating the direction of the key points, wherein the dimensionality of each key point feature vector is 128, and taking the first 120 best key points, the matrix of the descriptor is 120 x 128; and (3) clustering all key points by adopting a K-means algorithm, wherein the clustering number is 10, and each picture is represented as a bag-of-words model, so that each picture can be uniformly represented by a vector with the size of 10 x 1. And acquiring 123 hole images in different forms, wherein the ratio of the training set to the test set is 2:1, the training set comprises 82 hole images and the test set comprises 41 hole images.
The training set and its labels are sent to the SVM to train a supervised learning classifier, and the trained classifier is used to classify the test set and output the classification result, as shown in fig. 18.
The research of the invention has important reference value in the aspects of extracting the three-dimensional GPR hole characteristics and realizing the interpretation of the hole form. After the problem is solved, the problem related to the detection of the cavity diseases in the road structure in the real scene can be easily solved, which is the foundation for implementing and optimizing the compaction of the subsequent large road condition detection system.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, but any modifications or equivalent variations made according to the technical spirit of the present invention are within the scope of the present invention as claimed.

Claims (10)

1. The road cavity form classification method based on the three-dimensional GPR forward modeling technology is characterized by comprising the following steps of: the method comprises the following steps:
s1: three-dimensional GPR road hole modeling and three-dimensional GPR data acquisition,
adopting GprMax3D to build a three-dimensional road cavity simulation model and acquiring 3DGPR cavity data;
s2: extracting images in different directions of the cavity form,
and acquiring a B-scan image set and a C-scan image set along xoy and xoz planes respectively;
s3: the combined characterization of the shape of the hollow cavity,
arranging the B-scan images and the C-scan images in sequence, and combining and splicing the B-scan images and the C-scan images into a new two-dimensional image;
s4: the SIFT characteristics are extracted, and the method comprises the following steps of,
extracting scale invariant feature transform features (SIFT features) of the new combined image;
s5: the training test of the SVM classifier is performed,
carrying out K-means clustering and bag-of-word model conversion on the extracted SIFT features in sequence, then obtaining a new image vector, and using the new image vector as the input of an SVM classifier to realize the training and testing of the SVM classifier;
s6: the result of the classification of the shape of the cavity,
and after training and testing of the SVM classifier, outputting a classification result of the cavity form.
2. The road hole shape classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 1, wherein: in step S1: the three-dimensional road cavity simulation model comprises five layers, namely a first air layer (1), a second asphalt layer (2), a third concrete layer (3), a fourth gravel layer (4) and a fifth road base layer (5).
3. The road hole shape classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 2, wherein: the road base layer (5) simulates random medium to establish a soil layer model and soil dispersion materialThe definition is as follows: the soil has a sand content of 50% and a clay content of 50%, and the sand density is 2.66g/cm 3 Clay bulk density of 2g/cm 3 And a volume water content ranging from 0.001 to 0.25, the above materials being randomly distributed in a volume of 2X 1.2X 1m 3 On the model of (2).
4. The road void morphology classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 1, wherein: the cavity form includes spherical cavity, rectangle cavity, cylindrical cavity and irregular cavity, spherical cavity, rectangle cavity and cylindrical cavity have smooth surface, irregular cavity sets up to unsmooth surface.
5. The road void morphology classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 1, wherein: the three-dimensional GPR system parameters comprise a spatial resolution parameter of 0.01m, a time window parameter of 14ns, an initial transmitting antenna coordinate parameter of (0.45m, 1.0m and 0.0m), an initial receiving antenna coordinate parameter of (0.35m, 1.0m and 0.0m), an antenna stepping distance parameter of (0.01m, 0m and 0m), a measuring point number parameter of 100, an excitation signal type of Ricker and an excitation signal frequency parameter of 800 MHz.
6. The road hole shape classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 1, wherein: in step S2: extracting two-dimensional data of a two-dimensional section diagram B-scan and a horizontal section diagram C-scan from a three-dimensional structure S of GPR hole data, and performing shape recognition of the underground hole by using combined data of the two-dimensional section diagram B-scan and the horizontal section diagram C-scan:
1) two-dimensional data are sequentially extracted from S along a plane parallel to xoz, and a plurality of section images are obtained by imaging
S 1 ＝{B 1 ,B 2 ,B 3 ,…,B n }，n＝l y Δ y, wherein B n Represents each B-scan, S 1 Then it is a set of n B-scans, l y For the y-axis length of the model space, Δ y is the space step in the y-axis direction, and in addition, in the extractionAfter B-scan is obtained, preprocessing operations such as direct wave removal, data standardization, depth and speed change correction and the like are carried out on each image, signals and background noise are removed, the B-scan is extracted at equal intervals and displayed in a stacking mode;
2) two-dimensional data is extracted from S along the plane parallel to xoy, and a plurality of horizontal sectional images are obtained through imaging
S 2 ＝{C 1 ,C 2 ,C 3 ,…,C m }，m＝l z ,/Δ z, wherein C m Represents each C-scan, S 2 Is a set of m horizontal slices, l z To model the spatial z-axis length, Δ z is the spatial step in the z-axis direction.
7. The road hole shape classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 1 or 6, wherein: in step S3: fusing a plurality of B-scans and C-scans images, and screening 8B-scans { B ] from each cavity model 1 ,B 2 ,B 3 ,B 4 ,B 5 ,B 6 ,B 7 ,B 8 And 12C-scans { C } 1 ,C 2 ,C 3 ,C 4 ,C 5 ,C 6 ,C 7 ,C 8 ,C 9 ,C 10 ,C 11 ,C 12 And are arranged in sequence to form a new two-dimensional image, the image set I is expressed as formula (1), each three-dimensional cavity is finally converted into an image with fused information characteristics,
Figure FDA0003636474590000021
8. the road hole shape classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 1, wherein: in step S4: searching key points by adopting an SIFT algorithm and calculating the direction of the key points, wherein the dimensionality of a feature vector of each key point is 128, taking the first 120 key points, the matrix of a descriptor is 120 x 128, clustering is carried out on all the key points by adopting a K-means algorithm, the clustering quantity is 10, each picture is represented as a bag-of-words model, each picture can be uniformly represented by a vector with the size of 10 x 1, 123 hole images with different forms are obtained in total, the proportion of a training set to a testing set is 2:1, 82 training sets and 41 testing sets are obtained.
9. The road hole shape classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 1 or 8, wherein: in step S5: and sending the training set and the labels thereof into the SVM to train and test a supervised learning classifier.
10. The road hole shape classification method based on the three-dimensional GPR forward modeling technology as claimed in claim 9, wherein: and classifying the test set by using the classifier trained in the step S5, and outputting a classification result.
CN202210503794.8A 2022-05-10 2022-05-10 Road cavity form classification method based on three-dimensional GPR forward modeling technology Pending CN114998645A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210503794.8A CN114998645A (en) 2022-05-10 2022-05-10 Road cavity form classification method based on three-dimensional GPR forward modeling technology

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210503794.8A CN114998645A (en) 2022-05-10 2022-05-10 Road cavity form classification method based on three-dimensional GPR forward modeling technology

Publications (1)

Publication Number Publication Date
CN114998645A true CN114998645A (en) 2022-09-02

Family

ID=83025338

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210503794.8A Pending CN114998645A (en) 2022-05-10 2022-05-10 Road cavity form classification method based on three-dimensional GPR forward modeling technology

Country Status (1)

Country Link
CN (1) CN114998645A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117079145A (en) * 2023-10-17 2023-11-17 深圳市城市交通规划设计研究中心股份有限公司 Comprehensive road condition evaluation method, electronic equipment and storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117079145A (en) * 2023-10-17 2023-11-17 深圳市城市交通规划设计研究中心股份有限公司 Comprehensive road condition evaluation method, electronic equipment and storage medium
CN117079145B (en) * 2023-10-17 2024-03-26 深圳市城市交通规划设计研究中心股份有限公司 Comprehensive road condition evaluation method, electronic equipment and storage medium

Similar Documents

Publication Publication Date Title
CN107621626B (en) Radar signal Railway Roadbed detection method based on depth convolutional neural networks
CN112462346B (en) Ground penetrating radar subgrade disease target detection method based on convolutional neural network
Tong et al. Innovative method for recognizing subgrade defects based on a convolutional neural network
Kim et al. Deep learning-based underground object detection for urban road pavement
CN113256562B (en) Road underground hidden danger detection method and system based on radar images and artificial intelligence
CN102622607B (en) Remote sensing image classification method based on multi-feature fusion
CN114296075B (en) Ground penetrating radar image artificial intelligence identification method and device
CN110706211A (en) Convolutional neural network-based real-time detection method for railway roadbed disease radar map
Guldur Laser-based structural sensing and surface damage detection
CN104463248A (en) High-resolution remote sensing image airplane detecting method based on high-level feature extraction of depth boltzmann machine
JP2018041178A (en) Learning data generation method and target space state recognition method employing the same
Wang et al. Morphological approach to quantifying soil cracks: Application to dynamic crack patterns during wetting‐drying cycles
Li et al. Filtering airborne lidar data by modified white top-hat transform with directional edge constraints
Gong et al. Research on GPR image recognition based on deep learning
CN114998645A (en) Road cavity form classification method based on three-dimensional GPR forward modeling technology
Lehtomäki et al. Performance analysis of a pole and tree trunk detection method for mobile laser scanning data
Gui et al. Object-based crack detection and attribute extraction from laser-scanning 3D profile data
CN115616674A (en) Ground penetrating radar training data set simulation amplification and road nondestructive testing method and system
CN116203559A (en) Intelligent recognition and early warning system and method for underground rock and soil disease body
CN110019594A (en) A kind of geologic datebase method for building up based on numerical simulation
Li et al. MV-GPRNet: multi-view subsurface defect detection network for airport runway inspection based on GPR
Zahs et al. Classification of structural building damage grades from multi-temporal photogrammetric point clouds using a machine learning model trained on virtual laser scanning data
Xiong et al. Automatic detection and location of pavement internal distresses from ground penetrating radar images based on deep learning
CN105844707A (en) Method for automatically extracting ancient city wall data from LiDAR point cloud data based on cross section of city wall
Huang et al. Research on void signal recognition algorithm of 3D ground-penetrating radar based on the digital image

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination