Elevating point-based object detection in UAVs: A deep learning method with altitude fusion
[ 1 ] Wydział Automatyki, Robotyki i Elektrotechniki, Politechnika Poznańska | [ 2 ] Instytut Robotyki i Inteligencji Maszynowej, Wydział Automatyki, Robotyki i Elektrotechniki, Politechnika Poznańska | [ S ] student | [ SzD ] doctoral school student | [ P ] employee
[2.2] Automation, electronics, electrical engineering and space technology
2024
chapter in monograph / paper
english
- remote sensing
- deep learning
- GSD and altitude fusion
- data fusion
EN Recent advancements in computer vision and deep learning have revolutionised remote sensing. An important challenge lies in detecting small objects like individuals in crowds from low-altitude drone-captured images, which are problematic due to scale variations. While existing research addresses the scale challenges, not enough focus has been put on the exploitation of altitude data from UAV sensors. This paper proposes three deep learning-based methods to fuse altitude and Ground Sampling Distance (GSD) information, which enhance input images with additional features and showcase significant improvements in point-oriented object detection on our custom dataset.
publisher's website
final author's version
at the time of publication