Dissertation > Excellent graduate degree dissertation topics show
Spatial Uncertainty in Stereo Visual Odometry
Author: OuYangLiu
Tutor: LiuJiLin
School: Zhejiang University
Course: Information and Communication Engineering
Keywords: Stereo vision odometer Uncertainty Analysis Image feature points Motion Estimation Random sampling consistency test
CLC: TP391.41
Type: Master's thesis
Year: 2012
Downloads: 33
Quote: 0
Read: Download Dissertation
Abstract
From the point of view of the uncertainty evaluation of the performance of the visual odometry, as well as how to take advantage of the uncertainty of knowledge to improve the accuracy of the odometer, and are subject worthy of study. Based stereo vision odometer rigorous uncertainty analysis and demonstrate some core algorithm, combined with the uncertainty of the related information on the odometer improved algorithm to improve the accuracy of the odometer. The main content: paper first introduces the basic principle of stereo vision and this article is to study the threedimensional visual mileage calculation process, including the extraction of feature points, feature matching and tracking, least squares motion estimation and RANSAC removing Misassignments point . The paper studied the image feature points of uncertainty. The article summed up the general framework of the scaleinvariant feature extraction algorithm, given the variance of the feature point extraction transfer formula based on the framework, and an example to the SIFT verification of the simulation, and compared with existing methods. Motion estimation is a key issue for visual odometry. Image feature points of uncertainty first passed through triangulation formula to the threedimensional point of the threedimensional point positioning uncertainty the article based linear variance of the derivative nonlinear methods of delivery methods and tasteless transform this problem were solving the derivation of matrix perturbation theory is applied to solve the eigenvector derivative knowledge. Single movement parameter passing not enough to solve the problem of the uncertainty of the odometer, the article goes on to consider the uncertainty about the the RANSAC algorithm estimated parameters. This paper proposes a new algorithm based on Bayesian theory used to estimate such a huge amount of data uncertainty nondifferentiable pass, and achieved good results in the experimental verification. According to the gradual accumulation of movement, derived odometer positioning uncertainty formula, and proved the effectiveness of the algorithm in the actual experiment. 4. Combined with the uncertainty of the feature points, the article proposes a combination of the uncertainty of the image feature points maximum likelihood estimation method optimized single step motion estimation; proposed a partial binocular bundle adjustment of visual The odometer result further optimization. A large number of experiments carried out in many different environments, including indoor imitation lunar surface environment, grass, largescale campus environment, and highprecision total station acquisition real data to test the results of the visual odometry.

Related Dissertations
 Optimizing and Realising Research on Vedio Compression in TV Guidance System,TN919.81
 Research on Rotational Image Stabilization Algorithms for Electronic Image Stabilization System,TP391.41
 Study on Side Information Generation in Distributed Video Coding,TN919.81
 Research on Parallel Computation of H.264 Based on CUDA,TN919.81
 Study of Adaptive Motion Estimation in H.264/AVC,TN919.81
 Metrological Standard Device of Heavy Direct Current,TM835.2
 Study on Uncertainty of the Rocksoil Thermal Response Test in GroundCoupled Heat Pump System,TU831
 H.264 Motion Estimation Algorithm and Application of hierarchical coding,TN919.81
 Spectrum Information hyperspectral imaging target detection technique,TP391.41
 Head and shoulders image automatic segmentation of video,TP391.41
 Equal thickness interferometer for largesize digital measurement of flatness,TH744.3
 Passenger test equipment procurement project's economic feasibility study,F426.471
 Ultrahigh definition video coding based on synchronous hierarchical motion estimation architecture design,TN919.81
 OpenCVbased digital image stabilization system to achieve,TP391.41
 Frame rate upconversion algorithm,TP391.41
 Research on Moving Object Detected Technique for PTZ Cameras,TP391.41
 An improved BPM motion estimation algorithm,TP391.41
 The H.264 Inter block matching algorithm with encoder DSP implementation,TN919.81
 Research on H.264 Interframe Prediction Algorithm,TN919.81
 Research and Design on VLSI Architecture Supporting MBAFF Motion Estimation Engine,TN47
CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile
