Rychlá detekce beztexturových 3D objektů

Abstract

The goal of this thesis is to describe implementation and steps of detection algorithm. Purpose of the algorithm is to localize and detect textureless 3D objects in scenes and add label, which object was found. Inputs of the program are templates of objects, which are captured from all sides by different angles. The first phase is the preparation and is based on the input images. All prepared data are saved afterward, because this phase takes lots of time. The following step is the detection phase, which starts by loading prepared data. The main advantage of those prepared data is possibility to run detection phase repeatedly without negative time impact.

The implementation of the thesis is made in C++ programming language with OpenCV and OpenMP libraries. OpenCV is open–source computer vision library and contains functions and methods to process digital images and matrix. Library named OpenMP is used for parallelization of implementation and is already present in the development environment Visual Studio.