CN-115755193-B - Pavement structure internal disease identification method

CN115755193BCN 115755193 BCN115755193 BCN 115755193BCN-115755193-B

Abstract

The invention discloses a pavement structure internal disease identification method which comprises the steps of constructing a disease identification data set, acquiring a pavement disease type set and a formation mechanism based on the disease identification data set, acquiring a pavement disease map, identifying disease positions in the pavement disease map, determining disease types in the disease positions based on the pavement disease type set, identifying water diseases in the pavement disease map based on deep learning, and identifying internal cavities of a pavement based on a ground penetrating radar. According to the invention, through researching the recognition mechanism of asphalt road surface diseases based on computer vision and the recognition mechanism of road surface internal defects based on ground penetrating radar, a multi-technical-means road surface structure disease sensing and analyzing method is provided, and the road surface disease sensing precision is improved.

Inventors

LUO JUNHUI
HUANG XIAOFENG
LIU HONGYAN
CHEN DADI
ZHANG QIUCHEN
CHEN JIANGCAI
TANG HAO

Assignees

广西北投交通养护科技集团有限公司
广西威航道路工程有限公司

Dates

Publication Date: 20260505
Application Date: 20221026

Claims (6)

1. The method for identifying the internal diseases of the pavement structure is characterized by comprising the following steps of: constructing a disease identification data set and acquiring a disease type set and a formation mechanism of the pavement disease based on the disease identification data set; Collecting a pavement defect map, identifying a defect position in the pavement defect map, and determining the defect type in the defect position based on an obtained defect type set; Identifying water diseases in the pavement disease map based on deep learning and identifying internal cavities of the pavement based on ground penetrating radar; the process for identifying the water damage in the pavement damage map based on deep learning comprises the following steps: acquiring a response rule of the electromagnetic wave in the water disease area based on a finite time domain difference method; extracting features of the pavement disease map to obtain a water disease data set; performing water disease identification based on a two-dimensional convolutional neural network and the water disease data set; the process for identifying the internal cavity of the pavement based on the ground penetrating radar comprises the following steps: constructing air-filling and water-filling cavity models of different structure layer positions based on a finite time domain difference method; Acquiring a plurality of hierarchical structure images of the interior of the pavement based on the ground penetrating radar and the air-filled water-filled cavity model; Performing offset imaging processing on a plurality of hierarchical structure images to obtain radar wave simulation images; Forward modeling is carried out on the road cavity based on the radar wave simulation image, and a radar wave characteristic image is obtained; and based on the radar wave characteristic image and a YOLOv identification model, carrying out internal cavity identification.
2. The method for identifying internal road surface defects according to claim 1, wherein the process of constructing a defect identification data set and acquiring a pavement defect type set and a formation mechanism based on the defect identification data set comprises: Constructing a disease identification data set based on the formation cause of the road surface disease, the degradation effect of the road surface structure and the driving safety influence degree; And constructing a disease identification model based on the disease identification data set and the three-dimensional convolutional neural network, simulating a disease generation process, analyzing the road surface stress characteristics and driving safety change rules under the action of the disease, and the formation cause, the state type and the distribution range of road surface damage, and simulating the process that the damage defect in the road surface is developed into the disease.
3. The method for identifying a defective portion in a pavement structure according to claim 1, wherein the step of identifying a defective portion in the pavement defect map comprises: carrying out digital image enhancement, gray level treatment and binarization pretreatment on the pavement defect map; And constructing YOLOv5 identification model based on YOLOv network, and identifying the pretreated pavement disease map based on YOLOv identification model to obtain the disease position.
4. The method for identifying internal diseases of a pavement structure according to claim 3, wherein the YOLOv network comprises a backup part, a Neck part and a Head part, and the process for constructing the YOLOv identification model comprises the following steps: constructing a data set based on the pavement disease map, and dividing the data set into a test set, a training set and a verification set; Inputting the pavement disease image into a backup part of the YOLOv network to obtain disease feature images with different scales; Inputting disease feature graphs with different scales into the Neck part, sampling and feature fusion are carried out on the disease feature graphs, and tensor data with different scales are obtained; And inputting the tensor data into the Head part for gradient calculation, and verifying based on the verification set to obtain YOLOv identification model.
5. The method for identifying internal water damage in a pavement structure according to claim 1, wherein the process of identifying water damage based on a two-dimensional convolutional neural network and the water damage data set comprises: Constructing a classifier integrating a plurality of two-dimensional convolutional neural networks, and performing deep learning on the water disease data set; and carrying out weight control on each two-dimensional convolutional neural network classifier based on a voting scoring method, and outputting a water disease identification result.
6. The method for identifying internal diseases of a pavement structure according to claim 1, wherein the step of performing offset imaging processing on the plurality of hierarchical images comprises: acquiring wave speed ranges of electromagnetic waves in a plurality of hierarchical structure images based on a common dielectric constant; setting an offset parameter based on the wave speed range; extracting feature vectors of the hierarchical structure images based on a wavelet transformation method, and calculating wavelet entropy of each hierarchical structure image; taking the wave speed corresponding to the minimum wavelet entropy in the wave speed range as the optimal wave speed; and performing offset imaging processing on the plurality of hierarchical structure images based on the optimal wave speed and the offset parameter.

Description

Pavement structure internal disease identification method Technical Field The invention belongs to the field of road detection, and particularly relates to a method for identifying internal diseases of a pavement structure. Background In recent years, highway engineering construction continues to grow, due to unique climatic and hydrogeological topography conditions in the environment, early-stage built roads are subjected to increasingly serious early diseases, wherein road cracks are greatly generated due to uneven settlement of roadbeds, lower roadbed strength and the like, the continuous development of the cracks seriously affects the integrity of the road, service life of the road is reduced, and internal structural diseases of the road are difficult to accurately detect and measure, so that the internal structural diseases of the road are difficult to repair and treat in time, hollows generated under the long-term action of external load in the road are difficult to discover and treat in time, water leakage is generated at the hollows, various water diseases are further caused, the structural safety and service life of the road are seriously endangered, meanwhile, the difficulty of road management and maintenance is greatly improved, higher maintenance cost becomes a great burden for traffic management departments, and great negative influence is brought to managers and constructors. The composite asphalt pavement is in a typical road tunnel pavement form, fatigue cracking and reflection cracking are the most typical and widely occurring tunnel asphalt pavement defects, and the occurrence of the cracks greatly reduces the integrity of the pavement, has a larger influence on the bearing capacity of the pavement and seriously threatens the driving safety. At present, the traditional detection or perception technology for asphalt pavement cracks mainly comprises manual inspection, a rapid detection vehicle and various embedded sensors. Most areas still rely on manual inspection, but manual inspection is inefficient and has large errors. Along with the rapid development of image processing technology, laser scanning technology and the like, the rapid detection system for the pavement is rapid in development, the application of the rapid detection system obviously improves the single detection speed, but the identification precision of the prior art is low, and the requirements on the precise management and maintenance of the pavement cannot be met. Disclosure of Invention The invention aims to provide a pavement structure internal disease identification method for solving the problems in the prior art. In order to achieve the above object, the present invention provides a method for identifying internal diseases of a pavement structure, comprising the steps of: constructing a disease identification data set and acquiring a disease type set and a formation mechanism of the pavement disease based on the disease identification data set; Collecting a pavement defect map, identifying a defect position in the pavement defect map, and determining the defect type in the defect position based on an obtained defect type set; And identifying the water damage in the pavement damage map based on deep learning and identifying the internal cavity of the pavement based on the ground penetrating radar. Optionally, the process of constructing the disease identification data set and obtaining the pavement disease type set and the formation mechanism based on the disease identification data set includes: Constructing a disease identification data set based on the formation cause of the road surface disease, the degradation effect of the road surface structure and the driving safety influence degree; And constructing a disease identification model based on the disease identification data set and the three-dimensional convolutional neural network, simulating a disease generation process, analyzing the road surface stress characteristics and driving safety change rules under the action of the disease, and the formation cause, the state type and the distribution range of road surface damage, and simulating the process that the damage defect in the road surface is developed into the disease. Optionally, the identifying the disease position in the pavement disease map includes: carrying out digital image enhancement, gray level treatment and binarization pretreatment on the pavement defect map; And constructing YOLOv5 identification model based on YOLOv network, and identifying the pretreated pavement disease map based on YOLOv identification model to obtain the disease position. Optionally, the process of identifying the water damage in the pavement damage map based on deep learning includes: acquiring a response rule of the electromagnetic wave in the water disease area based on a finite time domain difference method; extracting features of the pavement disease map to obtain a water disease data set; and carrying out water disea