Surface Cracks Detection In Tunnel Concrete Lining Based On Multiple Feature Extraction And Percolation Model

With the rapid development of public transportation infrastructure,China has gradually become the country with large scale and most complex form of highway tunnel construction in the world.As the main disease of the tunnel concrete lining structure,crack is easy to form a vicious circle with the other areas,which can damage the durability and reliability of the tunnel.Digital image processing based crack rapid detection is aimed to overcome the defects and shortcomings of traditional instrument detection,and it is laid the theoretical foundation for the tunnel hazard rating and safety assessment.Meanwhile,it has a certain application value in preventing and curing concrete tunnel lining structure.This thesis intensive studied the surface cracks detection in tunnel concrete lining based on multiple feature extraction and percolation model.The status of surface cracks detection in tunnel concrete lining at home and abroad has been summarized.Compared with the existing crack detection technology,this thesis analysised the characteristics of surface cracks in tunnel concrete lining from the gray feature,edge feature,shape feature,etc.In this thesis,the proposed algorithm rapidly processed the practical engineering image,and effectively extracted the binary images of real cracks.The main research work is as follows:1.On the basis of distortion correction,smoothing and gray pretreatment of tunnel concrete lining surface images,this thesis studied the grid cell analysis.A multiple feature extraction based crack seed pre-extraction algorithm has been proposed.The multiple feature extraction in the grid and grid room has been improved to update the weight of pixel and the weight ratio of pixel extraction.The crack pixels have been pre-extracted,and the crack seed map which reduced the computational redundancy in late percolation has been generated.The detection efficiency has been improved.2.Combined with edge and shape feature extraction,this thesis studied second accelerated percolate and denoising algorithm based on percolation model.Aiming at the problem of fractured detection results and missed short crack,crack connection based on region extension has been improved.The connected dot pitch and the angle of the connected region have been modified.And the weight ratio of connection conditions has been set to connect and repair the crack of detection results.The connection region has been checked to improve the detection accuracy and reduce the false detections and missing detections.3.For the special background of inherent lining seam on tunnel concrete lining,this thesis proposed a lining seam elimination algorithm based on the feature unit.The lining seam has been segmentated to the feature unit-line of single pixel wide which can be minimal processed.According to the multiple feature extraction results of crack and lining seam,the marking criterion of unit-line has been designed to remove the background interference of lining seam and recover the cracks.And the more realistic cracks have been obtained.4.Surface cracks detection based on multiple feature extraction and percolation model have been studied.This thesis designed and implemented a system of surface cracks detection in tunnel concrete lining.The image is input to do real-time processing,and the binary images of real cracks can be output.The utilized goal of convenient,fast and easy has been achieved,and the requirement of fast and accurate crack detection has been satisfied.Finally,this research work has been summarized and the future prospect has been pointed out to open up the thinking of following topics.