Akademik Veri Yönetim Sistemi
INTERNATIONAL TECHNOLOGICAL SCIENCES AND DESIGN SYMPOSIUM (ITESDES 2022), Giresun, Türkiye, 2 - 05 Haziran 2022, ss.28
In this study, a signal processing
problem that includes different feature extraction approaches and produces
equivalent results is evaluated. The processed dataset is two-dimensional laser
measurements received from LIDAR sensor on a mobile robot platform navigated in
the Machine Vision Laboratory of the Science Institute. The dataset has been
filtered and represented with a more appropriate notation to provide a virtual
environment map. The resulting solution proposes a segment model that is close
to the real-world map. The segmentation process has been evaluated by
implementing different procedures. It has been shown that these procedures lead
to similar results. The specific features, advantages and disadvantages of
these methods are discussed. The algorithms that are widely applied and
therefore extensively tested and validated on such problem have been analyzed.
On the other hand, a less effective approach in this field (Random sample consensus)-(RANSAC) has been strengthened with a novel modification. Instead of
randomly selected contour points in the traditional RANSAC algorithm,
predefined subsets of the overall dataset are forced to be processed within the
boundaries determined by split and merge algorithm. The proposed solution
provides an important base for collision-free navigation of autonomous mobile
robot systems.