Design and Detection of Local Geometric Features for Deformable Marker Fields

augmented reality, deformable marker field, local geometric features

This paper sketched out the concept of the Honeycomb Marker Fields - a highly deformable and occlusion-tolerant marker field. We focused on the design of the Y-junctions, the visual image features which appear between the modules (cells) of the marker field. As the main contribution, we presented an algorithm for real-time and robust detection of the Y-junctions, and its experimental evaluation. The specialized Y-junction detection algorithm outperforms a general interest point detector (FAST) in the detection performance. The measurements thus lead to conclusion that for applications which can accept a marker field as a fiduciary pattern, the reliability and performance can be significantly improved. Based on the presented detector of Y-junctions and the preliminary HMF detector, we plan for the closest future to construct a robust and fast detector of the marker field and showcase it on selected applications. Thanks to the performance of the Y-junction detector, the algorithm is expected to be functional on today's ultramobile devices. The next step thus will be to implement it for some of the popular smartphone platforms.

A major limitation of contemporary fiduciary markers is that they are either very small (they try to represent a single point in the space) or they must be planar in order to be reasonably detectable. A deformable large-scale marker or marker field that would be efficiently detectable is the objective of this work. We propose a design of such a marker field - the Honeycomb Marker Field. It is composed of symmetric hexagons, whose triplets of modules meet at "Y-junctions". We present an efficient detector of these image features - the Y-junctions. Thanks to the specific appearance of these synthetic image features, the algorithm can be very efficient - it only visits a small fraction of the image pixels in order to detect the Y-junctions reliably. The experiments show that compared to a general feature point detector (FAST was tested), the specialized Y-junctions detector offers better detection reliability.