A Classification-Based Algorithm for Building 3D Maps of Environmental Objects

In this paper, a method for building a 3D map of some objects detected in an indoor environment is presented. The pecularity of the proposed algorithm is that it works with a simple consumer-grade webcam. With the webcam, pictures of the environment are taken. The proposed method first extracts the regions which may contain an object from the pictures. The regions are then classified to identify the objects, further their pose and height are estimated. A 3D map of the environment is finally reconstructed where icons roughly resembling the object categories are added to the 3D map at the estimated object position and with the estimated height. Regions of Interest (ROIs) extraction is performed using Haar-like algorithm. Before classification, the images containing the ROIs are processed to extract the edges of the objects. Non relevant edges are removed using a novel fuzzy technique. Object classification is performed with a pseudo2D-HMM algorithm. Experimental results are presented for an office environments.