The Research Of Real-time Based Width Measurement Technology Based On Binocular Stereo Panoramic Vision Sensor

With the development of Chinese urbanization,the pressure of population,energy,transportation,infrastructure and others is gradually increased In order topromote the development of urban construction,the conce。pt of''Digital City''W'asproposed''Digital City"is the three-dimensional description on city bymulti-resolution,multi-scale,multi-dimensional space-time and multi-species,style toachieve digital economy development and public serviceThe construction of fundamental database is the strategic foundation of"digitalcity",and the data collection and processing is the prerequisite on it However,thetraditional working pattern,not only requires a lot of manpower,material andfmancials,but also inefficient It is difficult to keep up with rapid changes in urbantransportation construction The Mobile Mapping System(Mobile Mapping System,MMS)provides a convenient and economic pattern,which can meet the requirementsof infrastructure database construction in intelligence transportationThe MMS is cutting-edge technology in mapping currently It is fitted in thevehicle global positioning system(GPS),video system(CCD),inertial navigationsystem(INS)or the dead reckoning system(DR)and other advanced sensors andequipment,as the vehicles moving with high-speed,the MMS can acquire the spatialdata and attribute data of roads,all these data are stored in the onboard computer inreal time or via processing to form a variety of useful data sets integrated systemHow-to use the CCD video system to capture rich image from both sides of theroads and surface features or to calculate spatial data,attribute data accurately in real-time to achieve real three-dimensional information system,this is a key issue oftheMMSThis selected article is based on the video system of binocular stereo pmaoramicvision sensor on width measurement application of the road The binocular stereopmaoramic vision sensor is designed vertically by two Omnidirectional vision senso~,this design can capture image,extract spatial data and attribute data with 360。rmageofroadThe thesis is based on omnidirectional vision sensors,and w-e also introduce theprinciple of binocular stereo vision in panoramic visual field In the panoramic vision,the recogn~ion algorithm on the road w-as studied and the binocular stereo pmaoramicvision system w-as analyzed comprehensively,including the calibration ofomnidirectional vision sensor,pmaoramic image capture and preprocessing,featurepoint identification and matching,the measurements of three-dimensional spatialinformationThe major w-ork ofthis thesis involves the followed aspects:1 The research on Mobile Mapping Platform is based on the stereo panoramicvision sensor We introduce the current Mobile Mapping Platform status andrepresentative LD2000-RM-based Mobile Mapping System briefly and propose theimprovement of the CCD visual system,then introduce the Omnidirectional visionand binocular stereo vision in the MMS 111e Omnidirectional vision sensor with 360。horizontal 120。vertical field ofvision can capture the large range ofroad images Ithas a single viewpoint imaging properties which can measure the spatial informationof image On this basis,w-e design a vertical structure of the stereo pmaoramic visionsensor(Binocular Omni-stereo Vision Sensor)The vertical structure of it meetsepipolar constraint of stereo matching,can compress image matching complexityfrom two-dimensional image retrieval into a very one-dimensional line It reduces thecomplexity of stereo matching greatly,while reducing the possibility of false matchesand improving the accuracy of stereo vision2 The road recognition algorithm under pmaoramic vision As the road imageunder panoramic vision is the distortion,which caused the inconvenience of road recognition In this thesis,we proposed the feature-based road detection algorithm,that is a combination detection algorithm based on regional characteristics and edgefeatures of the road in color image:First,we transform the color image from RGBcolor model into HSI color model and use the OTSU algorithm for adaptive thresholdsegmentation,SO that the rough road was detected；Next,using Canny algorithm,Hough algorithm detect the edge information of the road,and then update the road'Sdetail edge to get the accurate road area；Finally,on this basis,we can get thereference points of road width measurement In the whole of road detection,we notonly focus on the accuracy,but also take full account of Interference from complexenvironment,to improve the robustness ofthe algorithm3 The research of binocular stereo pmaoramic vision system based onomnidirectional vision sensor After introducing the calibration of the omnidirectionalvision sensors mad panoramic image correction method briefly,we focus on imagematching of binocular stereo vision system With using the epipolar constraintimplement t'ast stereo matching,we realize accuracy match of feature points by SIFTalgorithm and improved ASIFT algorithm in whole affine space,wide viewing maglerange So that we can overcome the Interference efl'ectively in stereo image matchingdue to camera rotation,translation,scaling,and illumination,affine mad perspectivechange,restrain the non-match,mismatch,match instability mad other issues causedfrom the environmental impact4 The analysis and research on accuracy and real-time of binocular visionsystem We introduce the principle of binocular parallax distance in the field ofbinocular stereo panoramic vision,to measure the coordinates of any object point inthree-dimensional spherical space,and then,according to the environment of road,wecontribute the road width measurement model to compute the road width Finally,weanalysis the accuracy and real-time of binocular vision system,weight the impactfactors on the measurement accuracy,mad propose improvement methodscomprehensively...