Open AccessDATA PREPARATION METHODS FOR RECOVERING THREE-DIMENSIONAL STRUCTURE OF THE SCENE
Author(s) -
Sergey Sergeyevich Taranenko,
Oleg Fedorovich Kovalev
Publication year - 2014
Publication title -
vestnik donskogo gosudarstvennogo tehničeskogo universiteta
Language(s) - English
Resource type - Journals
eISSN - 1992-6006
pISSN - 1992-5980
DOI - 10.12737/5717
Subject(s) - computer vision , artificial intelligence , computer science , sobel operator , filter (signal processing) , basis (linear algebra) , object (grammar) , wavelet , frame (networking) , noise (video) , pattern recognition (psychology) , edge detection , image (mathematics) , image processing , mathematics , telecommunications , geometry
The problem of data preparation to recover the three-dimensional structure of the scene including points, segments, straight lines, etc., using video obtained from a moving camera around a static scene is considered. Based on the analysis of the existing methods, the methodology and algorithm of the data selection and preparation for recovering the three-dimensional scene structure are provided. Canny filter and Sobel operator complemented with calculating the algorithm SURF (Speeded Up Robust Feature) descriptors is taken as a basis. To filter noise from frames, wavelet filters are proposed. This allows obtaining an object contour extraction algorithm in the frame, and its comparison with the object contour in the next frame, which is the input data for the majority of algorithms of the three-dimensional scene reconstruction.