Generating view models of monotone polyhedrons for visual identification. Presentation of results of the PhD thesis

Abstract

Main results the PhD dissertation are presented in which the goals were: to develop appropriate methods and algorithms in order to obtain a complete view representation of the object using the idea of the viewing sphere with the perspective projection, to compare and select the best algorithm for this task, and to show how the viewing models can be matched with objects extracted from range images. It was proved that it is possible to obtain the approximate view representation of a monotone polyhedron using the algorithm with the complexity of O(n2) and that the minimum number of matching view representations of a monotone polyhedron given with the data from range images is O(n2). The methods of generating view representations of monotone polyhedrons are investigated for the purpose of visual recognition systems and some of them are presented. View representations are used in many areas including identification tasks, e.g. as an element of visual channel in automation and robotics. The 3D view representation of objects can be constructed offline and saved in a database for further use in the recognition task.