The Algorithm Of Lining Seam Elimination In Tunnel Surface Images

With the rapid development of tunnel transportation infrastructure,tunnel complications have also followed.As the main issues of tunnels,cracks can badly influence the working areas of tunnels,thus making a vicious circle and seriously affecting the safety of tunnels.Using the technique of crack identification based on digital image processing,the defects and deficiencies of traditional crack detections made by humans and instruments can be overcome,and vital suggestions about tunnel issue evaluation and daily maintenance can be given at the same time.However,as for the crack detection on tunnel lining surface,existing pavement crack detection methods will meet great difficulty because of the inference of noises like lining seams,which have similar gray values and textures to cracks,thus pure cracks cannot be accurately extracted from tunnel lining surface images.The thesis researched on how to remove the interference caused by lining seams in tunnel crack images.Current studies of crack detection and lining seam elimination at home and abroad have been summarized.Comparative experiments of the existing line edge detection algorithms have been conducted in order to extract salient line segments at crack edges and lining seams edges in images.All line segments on the two kinds of edges were then classified according to the proposed rules.A lining seam elimination algorithm based on line segments' translation and expansion has been presented,and it can efficiently remove noises of lining seams and preserve real cracks as much as possible.The main works of the thesis are listed as follows:1.By studying the relevant references in domestic and international on linear detection and using the existing pavement crack detection algorithms to obtain a rough crack extraction from tunnel lining surface images,the thesis researched on how to extract salient line segments.For different line detection algorithms,the numbers of extracted line segments and the execution time of extracting process were quantitatively evaluated,and the local details of extracting images were carefully compared.After that,the core reasons caused the differences were analyzed in detail.The experimental results show that the extraction of the LSD is more efficient and accurate,and using the LSD can extract numerous and detailed line segments on the crack and lining seam edges in tunnel lining crack images.2.When the salient line segments extraction was completed,the thesis researched on how to classify these line segments,i.e.,to judge whether they were at the edges of cracks or at the edges of lining seams.Characteristics of straight line segments at two kinds of edges were summarized.Besides,the thesis designed a uniform k-divided angle region model to help finish the classification.The oblique angles of line segments were revised so that they can be in a determined range,and the accumulator was set to count the number of straight line segments in each sub-angle region.Combined with each line segment's length,oblique angle,and the value of the accumulator in the sub-angle region where the current line segment belongs,all straight line segments on the cracks or lining seams can be precisely classified thus the efficiency and accuracy of subsequent processing can be improved.3.After salient line segments classification,the thesis researched on how to quickly eliminate lining seams pixels and preserve crack pixels in tunnel lining surface images.An algorithm of eliminating lining seams in tunnel lining surface images via line segments' translation and expansion was proposed.By following the two new principles about the translation direction and expansion decision made for line segments and considering line segments classification results,the translating and expanding operations decomposed the overall structure of lining seams little by little,gradually processed lining seams pixels into background pixels.The experimental results show that the proposed algorithm greatly improved the accuracy of crack extraction on the tunnel lining surface,thus real cracks can be obtained.Finally,this thesis summarized the whole research and pointed out the potential ideas for future work.