DETECTING NATURAL OBJECTS BY MEANS OF 2D AND 3D SHAPE ANALYSIS

This paper describes two different approaches to detect natural objects based on their shape properties. In both cases it uses 2D snapshots of a natural scene, taken by a common digital camera. With a help of digitalized 2D images, shapes such as circles, can be identified with a help of Canny edge detector and Hough transformation, described in the first part of the paper. Canny edge detector takes a 2D image and converts it into its binarized version, revealing pixels that make up the edges of the objects. Edge pixels are then used by Hough transformation that detects prominent shapes. In the second part, the idea of 2D analysis is then taken and extended in to 3D space. It starts with 3D shape reconstruction, followed by a 3D Hough transform. 3D reconstruction enables us to transform 2D pixels into 3D voxels, which are then analysed by a 3D Hough transform.