A Method to Segment Maps from Different Modalities Using Free Space Layout MAORIS: Map of Ripples Segmentation

Abstract

How to divide floor plans or navigation maps into a semantic representation, such as rooms and corridors, is an important research question in fields such as human-robot interaction, place categorization, or semantic mapping. While most algorithms focus on segmenting robot built maps, those are not the only types of map a robot, or its user, can use. We have developed a method for segmenting maps from different modalities, focusing on robot built maps and hand-drawn sketch maps, and show better results than state of the art for both types.Our method segments the map by doing a convolution between the distance image of the map and a circular kernel, and grouping pixels of the same value. Segmentation is done by detecting ripple like patterns where pixel values vary quickly, and merging neighboring regions with similar values.We identify a flaw in segmentation evaluation metric used in recent works and propose a more consistent metric. We compare our results to ground-truth segmentations of maps from a publicly available dataset, on which we obtain a better Matthews correlation coefficient (MCC) than state of the art with 0.98 compared to 0.85 for a recent Voronoi-based segmentation method and 0.78 for the DuDe segmentation method. We also provide a dataset of sketches of an indoor environment, with two possible sets of ground truth segmentations, on which our method obtains a MCC of 0.82 against 0.40 for the Voronoibased segmentation method and 0.45 for DuDe.

Area Graph: Generation of Topological Maps using the Voronoi Diagram

“…Now follows an experiment to analyze the influence of the α value, using Matthew's correlation coefficient (MCC) [7] for segmentation measurement against a human-made ground truth. The range of MCC is from -1 to 1, where with the best segmentation MCC=1.…”

“…Mielle et al [7] segment maps by calculating distance image from them, in which each pixel has a pixel value that represents the size of the region it belongs to, i.e. the distance to its closest obstacle.…”

“…This experiment compares our algorithm with Mielle's MAORIS method [7] and the traditional Voronoi-based segmentation implemented by Bormann [6]. We used the best parameter described in their papers for the other two segmentation algorithms.…”

“…Already 1998 Thrun [4] proposed to use critical points along Voronoi Graphs for segmentation. Other, often similar segmentation algorithms The authors are with the School of Information Science and Technology, ShanghaiTech University Shanghai 201210, China, and also with the University of Chinese Academy of Sciences, Beijing 100049, China [houjw, yuanwj, soerensch]@shanghaitech.edu.cn have been presented since: [5] [6] [7]. In contrast to the mentions approaches, our algorithm is directly anchored on the topology of the Voronoi Graph, so it is separating areas only on junction points.…”

URSIM: Unique Regions for Sketch Map Interpretation and Matching

“…For segmentation of the sketch and metric maps, we use the MAORIS [26] algorithm. MAORIS' main advantage is its ability to segment maps from different modalities in a way similar to that of a human, reducing errors due to differences of representation or level of detail.…”