The Study On Calculating And Interpreting The Optical Flows Of Lines

It is a major topic in computer vision to recover the movement and structure of a 3-D moving object through the analyses of dynamic image sequences. The method based on optical flow fields is a powerful one to vision motion analysis.In this paper, we have researched the algorithm for calculating the optical flow based on color images, and have especially discussed the optical flow estimation techniques based on the feature lines. The paper also addresses 3-D recovery of an object from the optical flow fields of the feature lines.Firstly, the paper presents the conception of line coordinates to study the optical flow estimation techniques based on the feature line. We consider all kinds of plane line representations of a plane line, and a simple and practical line coordinate system is suggested. We also present a new 3-D line representation model. The new representation model is based on the geometry theory: the intersection of two arbitrary planes in the 3-D space is a line in a plane. One of the two arbitrary planes is formed by the origin and the projected line, which is called the inverse perspective projection plane, the other is perpendicular to the inverse perspective projection plane, and go through the 3-D line. The new representation model takes on the perfect, unique and simple to 2-D projection.Secondly, we point out the problems with previous straight-line extraction, and present a new algorithm for enhancing the accuracy of the parameter extraction of straight lines in a 2-D projection images. The new algorithm starts from an initial estimate which is based on the Hough transform, and a rectangular window is centered using the current line approximation, and a new line estimation is generated by making a total least squares fit through the pixels contained within the window. This is repeated until convergence is reached.Lastly, we have suggested a new technique which may recover the motion and structure parameters of a moving object by using of 21 optical flow lines based on the optical flow fields of the feature lineThis thesis is the project supported by Aeronautical Foundation Science ( No.99F53065) and Research Center of Measuring and Testing Technologies, and Control Engineering in Nanchang Institute of Aeronautical Technology(No.2001-15).