The team proposed a method to detect feature points and feature lines that are essential for simultaneous localization and mapping(SLAM) of a UAV using camera image information. For the brightness gradient direction angles at the edge point, we determined the point where the maximum rate of change within a certain neighborhood is greater than the threshold to the corner point (the feature point), and the nearest edge line to the feature line, and compared the detection performance and speed. Edge detection was done with Canny algorithm.
The proposed method is applied to stereo camera images of the EuRoc database and compared with previous studies.
First, we show that the corner points of the proposed method correspond to the feature points of the FAST method, and that the dependence on the threshold of the FAST method disappears when the edge is detected.
Then, we compared the previous edge-based corner point detection methods, CSS, PCD, and triangle theory, in terms of computational time, and compared them with ORB feature point detection, to show improved detection performance.
The results were published in the "Int. J. Advanced Networking and Applications"(Vol. 17, Issue 01 Pages 6761-6771(2025) ISSN:0975-0290) under the title of "A Method to Detect Corner Points and Feature Lines Based On Edges in Images in Visual SLAM"(https://www.ijana.in/papers/v1711-7.pdf).
