Research Of Pavement Damage Detection System Based On Binocular Vision

Road cracks detection plays an important role in the highway management andmaintenance. Using traditional manual way to detect road cracks cannot meet the needsof the quick development of highway management and maintenance. With the rapiddevelopment and popularization of photovoltaic technology and computer technology,the ways of road cracks detection based on photo-electric detection technology havebecome an inevitable trend. In this thesis paper, I mainly research the ways ofthree-dimensional pothole detection, make a systematic design and deeply researchvarious methods of calculating.This paper briefly describes the architecture of the three-dimensional potholedetection system, working principle of the three-dimensional pothole detection system,and the methods of choosing hardware and its reasons. In this paper, I summarize theimportant technologies to solve binocular vision imaging through analysis of domesticand international visual demarcation of the models and methods. Relying on the basis oftraditional calibration, I propose a new method of camera calibration based on thecircular standard targets and use RAC two-step calibration method to demarcate camera.According to the damaged image of potholes, I find various calculation methodsusing digital image processing technology including preprocessing the image of roadappropriately with gray-scale image processing technology and image filteringtechnology; reviewing the calculation way of edge extraction, choosing Canny templateas edge detection template and positioning to the image edge accurately; usingmathematical morphology to calculate the erosion and dilation of binary image.Finally, I describe the specific steps and processes of the implementation ofthree-dimensional road detection and reconfigurable systems, according to matchingprinciple to find matched characteristic point, establishing correspondence betweenworld coordinate and image coordinate, reconstructing three-dimensional shape ofdamaged pothole using the basic principles of binocular stereo vision. In the process ofthree-dimensional detection, I analysis the system error and propose a way to check outthe error. The experimental results show that three-dimensional road cracks detectionand reconstruction methods proposed in my thesis paper are very simple, fast and highlyaccurate.