Road Crack Detection Based On The Gravity Model And Mean-Shift

From the late 70 s of last century,road disease detection system firstly started research and development abroad,it has always been a hot Issue in academic field.It is the focus of attention whether a series of diseases on the road can be accurately identified,classified and counted using fully automated system.In recent years,a large number of domestic and foreign scholars engaged in the research and development work in this field.Relying on the rapid development of science and technology,and continuous practice and innovation,it has achieved excellent academic results in improving data collection,storing and processing equipment and improving the automated data processing algorithms.But the most important thing in the field is the real-time automated detection to the disaster such as The road Line crack,Fracture,broken,groove.The traditional approach is mainly dependent on hardware technology innovation to solve the problem of disease segmentation,such as improved data acquisition equipment and lighting equipment.After searching lots of literature,the traditional ideas have been changed,through the data processing of the road disease images,the algorithm with better robustness is used to deal with the road disease images with low resolution and fuzzy cracks,and the diseased areas in the images are precisely divided.In this thesis,we aim at detection and segmentation of road crack using low contrast image.The main research is briefly concluded as follows.(1)Illumination compensation algorithm is studied,which can eliminate the influence of light,providing convenience for target detection.(2)When preprocessing the road image,this thesis put forward an algorithm named improved bilateral filtering,which can make noise smooth as while as keeping outline of the crack.this algorithm can get a good result in filtering the road crack image.(3)When extracting the target,this thesis use an algorithm to find the scope of the cracks quickly using template matching method.After the test we know this algorithm can meet the requirements of detection.(4)When segmenting the crack accurately,this thesis study lots of algorithm about super pixel algorithm,and then,use an algorithm named shrinking contours algorithm with the idea of Mean-shift.And the final result can fit the requirements of detection.