Sharing landmark information using mixture of Gaussian terrain spatiograms

Abstract

Abstract-In this paper we evaluate the use of a novel spatial histogram called the terrain spatiogram as a common representation for exchanging landmark information between robots working as a team to map an area. Individual robots use range sensors to provide the spatial dimension of the spatiogram and video for the image dimension. We have previously shown that terrain spatiograms can be shared between robots in a heterogeneous team to recognize landmarks and to fuse observations from multiple sensors or multiple platforms.A terrain spatiogram using a mixture of Gaussians (MOG) model is introduced and a corresponding normalized spatiogram similarity measure defined. Two methods to generate a MOG terrain spatiogram are presented and compared experimentally using indoor and outdoor landmark information transferred between two different models of robot equipped with differently configured stereocameras.

Selection and recognition of landmarks using terrain spatiograms

“…The strong diagonal band (of 7 3×3 submatrices) shows that different poses of a landmark are well recognized and well distinguished from other landmarks -despite the somewhat similar shape and color of the three garbage bins (Figure 6 (a), 6(c) and 6(h)) for example. This result from automatic landmark selection by the LSA reinforces our previous results for manually selected landmarks, documented in [4] [5].…”

“…1 In [5] we defined TSGs that employ a mixture of Gaussians spatial distribution and the corresponding normalized comparison function, and demonstrated how this could be 1 It can be easily verified that ρ(h,h)=1.…”

“…A spatiogram using terrain rather than image spatial information is called a terrain spatiogram (or TSG). We have shown that the TSG is an effective approach to sharing information between robot platforms [4], combining multiple views of a landmark [5], and detecting and filtering landmark occlusions [6] when the input images are windowed to manually selected landmarks.…”

Detection and filtering of landmark occlusions using Terrain Spatiograms

“…Another representation that combines depth and image information is Birchfield and Rangarajan [1] information is related in a rather complication fashion to the scene spatial information. We proposed an extension to the spatiogram, called the Terrain Spatiogram [7] [8], in which the image spatial information is replaced by terrain spatial information. In [7] we presented experimental results for mutual landmark recognition on two different model robots equipped with different stereocameras, and with terrain spatiograms collected on one robot being used on the other.…”

“…In this paper, we investigate the problems of filtering landmark occlusions using a combined image and terrain spatial representation, Terrain Spatiograms, proposed in [7] [8]. Occlusion can be defined naturally in terms of depthrelated visibility constraints.…”