Cluster-Agnostic Line Detection for Computer Vision
Loading...
Authors
Said, Charles
Date of Issue
2025
Type
Thesis
Language
en_US
Subject Keywords
Other Titles
Abstract
A core challenge for autonomous robots is implementing computer vision techniques to detect navigable spaces around the robot and to differentiate navigable spaces from non-navigable ones. In this project, we develop a line detection algorithm using clusters of points interpreted as lines, which indicate the edges of the navigable space for our robot. This algorithm, combined with open-source keypoint extraction techniques, theoretically enables a robot to see the edges of its navigable space, a critical first step in pathfinding. Importantly, this algorithm is durable even when the number of lines in the original image is unknown. The algorithm is also relatively lightweight, relying only on a video stream and no other sensors.