Abstract

A 2D edge analysis scheme has been formulated based on (i) the wavelet transform; (ii) a data structure/object model named HSOBB with multi-resolution support; and (iii) multiple scale significance testing. It quickly isolates contour pieces that may belong to areas under risk. In real time applications, where the test for boundary changes is made in each time step, this is very important for studying the evolution of geotechnical phenomena and handling smooth curves as well as polygonal approximations. HSOBB introduces some noise filtering as well. It is a good option to more classical procedures, since it is able of dealing with a wide range of practical problems. Edges or abrupt jumps in contours may be enhanced by the use of different types of wavelet transform. Additional performance enhancements can be accomplished if noise is not assumed to be Gaussian, since commonly used detectors may have other types of noise ([12], [14], [15]), and may be subject to nonlinear distortions. An algorithm to detect exact interference for general and polygonal models was introduced. It pre-computes a hierarchical model representation using tight fitting OBB trees. At runtime, it goes across two such trees and tries to find overlaps among OBBs based on the separating axis theorem. In particular, it can robustly and accurately detect contacts between large complex geometries.