Semantic Analysis of 2D Maps With a Metric-Topological Approach

From ISLAB/CAISR
Title Semantic Analysis of 2D Maps With a Metric-Topological Approach
Summary Semantic Analysis of 2D Maps With a Metric-Topological Approach.
Keywords semantic labeling, place labeling, segmentation, pattern recognition, classification, 2D maps
TimeFrame Start: February 2014, End: June 2014
References Rottmann, Axel, et al. "Semantic place classification of indoor environments with mobile robots using boosting." AAAI. Vol. 5. 2005.

Liu, Ziyuan, and Georg von Wichert. "Extracting semantic indoor maps from occupancy grids." Robotics and Autonomous Systems (2013).

Schroter, Derik, Michael Beetz, and J-S. Gutmann. "Rg mapping: Learning compact and structured 2d line maps of indoor environments." Robot and Human Interactive Communication, 2002. Proceedings. 11th IEEE International Workshop on. IEEE, 2002.

Prerequisites Image analysis, programming skill, machine learning
Author
Supervisor Björn Åstrand, Saeed Gholami Shahbandi
Level Master
Status Internal Draft

Generate PDF template

A fundamental ingredient for semantic labeling is a reliable method for determining relevant spatial features of an environment and a proper way of describing the context through pattern classification. Adaptive Grid detects arbitrary dominant orientations in, fits corresponding line features with tunable resolution, and extracts topological information. This method is developed for the case of a maps with straight lines and occupancy informations. The goal of this project is to generalize the method for arbitrary feature fitting and develop a pattern classification for a more general map (like an aerial image of an structured area) other than just occupancy map. Research Question: How to employ pattern classification as region descriptor and developing a general scheme for semantic labeling. Work package P1: generalization of feature fitting, both in nature and shape. Features to be fitted and representing the structure of the environment do not have to be necessarily walls and straight light. Work package 2: analysis of the cells represented by Adaptive Grid (pattern classification) Work package 3: semantic label development Deliverable: an implementation and demonstration of the developed method for a general semantic analysis based on Adaptive Grid. Result of the method shall be demonstrated on other types of maps/images other than occupancy grid maps.