Stereo Vision For Asphalt-pavement Deformation Detecting Based On Smart Phone And Cloud Server

Pavement deformation damages have badly effect on road security,and it always is the focus object on road maintain.But current pavement deformation detection systems have issues such as high cost,bad portability and stand-alone system.Stereo vision method is widely applied on 3D reconstructios and virtual reality and extended to pavement deformation detection due to its perfect theories.It becomes more and more popular in the field of pavement deformation detection recently,but there are some problems in features matching,point cloud geometry calculation and 3D assessment indexs aspects.For solving the problems mentioned above in pavement deformation detection,this thesis take strength from stereo vision measuring method to do some research on asphalt-pavement deformation detection,developing a system for asphalt-pavement deformation based on android system smart phone and cloud service platform.This system can detect asphalt-pavement deformation with a portable and connected work style and greatly reduce the cost of detecting asphalt-pavement deformation.Finally,detection of asphalt-pavement deformation is achieved via this system.The research work done in this thesis could be listed as following:(1)The principle of stereo vision measurement is introduced,covering basic projective geometry,camera geometry and two-view geometry three aspects.More specifically,the illustration of the basic knowledge,like pinhole imaging,camera&lens relations,project transformation,fundamental matrix and essential matrix,provider theories for stereo vision study.(2)An insight introduction of stereo vision algorithms is gave out.Not only a detail introduce of structure from motion(SFM)algorithm is made,including features detection,features matching,camera pose estimation,scene structure and bundle adjustment,but also a better introduce of patch multi-vision stereo(PMVS)algorithm is done,including basic models,initial point cloud generation,patch expansion and outlier filtering four aspects.A dense 3D reconstruction program is achieved by using open source software library such as OpenCV,SSBA and PMVS.(3)Assessment index and detection methods of pavement deformation are studied.The mechanisms of the four common pavement deformation such as rut,sag,pothole and bump are analyzed and current deformation assessment indexes are summarized.After doing some research on pavement deformation detection and assessment standards,a stereoscopic pavement deformation assessment method for 3D point cloud of pavement deformation is proposed.(4)The demand of asphalt-pavement deformation detection is analyzed.Taking portability and operation efficiency into consideration,a cloud service system for pavement deformation is designed.With LAMP server architecture,HTTP transmission,Open GL ES graphic library,Glide image loading library,Camera2 API and multi-mode positioning technologies,a system for pavement deformation based on android system smart phone and cloud service platform is developed.This system module functions,including data collection,uploading,processing and result preview,are achieved simply.Importantly,it will further promote pavement deformation detection system development in simplification and interconnection.(5)Based on the proposed pavement deformation assessment method,stereo vision measurement method and point cloud processing method,an asphalt-pavement detection experiment carried out on Jiefang Road and Beijing Road of Xuzhou city central.Four common pavement deformations are detected via stereo method and results show that the stereo vision measurement method could achieve an accuracy detection of pavement deformation index when compared with manual measurement.