Topology Graph Pruning for Optical Mapping Methods using Edge Betweenness Centrality
Armağan Elibol, Nak Young Chong · 2019
Optical mapping is one of the most widely used application areas of low-cost robotic platforms. These platforms are in favor as they are relatively easy to use, to operate and to maintain. Acquired optical data (in the form of video and/or image) are valuable sources of information for both online (e.g., navigation, localization, mapping, and others) and offline processes (scientific interpretations, change detection, mapping, and others). The amount of data acquired has been continuously growing thanks to the emerging capabilities of mobile platforms in terms of autonomy allowing longer surveying time. This increases the need for fast and efficient methods to process the obtained data. Creating optical 2D maps from acquired data is composed of mainly image matching, trajectory estimation (Global Alignment (GA)) and image blending steps. In this paper, we discuss the usage of Edge Betweenness Centrality (EBC) concept to reduce the total number of overlapping image pairs to be used in the GA step. EBC allows selecting the image pairs that play a relatively key role in the topology graph. We also discuss the usage of graph energy as a decision criterion during image mosaicing iterations. We present experiments with several datasets to show the performance of the proposed method.