Object Contour Estimation using Stereo Vision

Abstract

In this thesis we will explain how we can use stereo vision to estimate the distance between the object and camera By using back projection method to 3-D space we recover existing two images to analyze the contour of the object Stereo matching is the most important step of stereo vision as it will determine the accuracy of our experimental result- estimated distance between the camera and the object Besides we will also be using unsharp mask filter to sharpen images and altering window sizes to find the best way to improve the accuracy and speed of stereo matching in our experiment In the first part of thesis we will explore how SURF algorithm is used to extract feature points and how use methods to estimate the fundamental matrix and the corresponding relationship between two images Next we use epipolar geometry to rectify two images and prove that the epipolar lines of rectified images are parallel and have the same height Last but not at least we use FNCC to match two images This experimental results demonstrate that the smaller the window is the more errors stereo matching occur and vice versa the bigger the window is the more actual result you get- sharpen images will be clearer than the original images

Date of Award	2014 Jan 20
Original language	English
Supervisor	Ta-Chung Wang (Supervisor)