Projects:Student:Computer Vision:Stitching without breaking objects
Title
Stitching without breaking objects.
Description
When having a camera sensor mounted on a mobile robot, taking pictures of the ground while the robot moves one might want to stitch those images together. If the displacement of the robot/camera from successive images is unknown this stitching process requires some overlap of images. This also means that single objects are visible in several images. The task of extracting single objects could be done in a stitched images but this would mean that part of the object might origin from different images even though the needed overlap of images could result in an object being totally included in one of the images. Extracting an object from a single image is expected to result in a higher quality than having to merge images taken from different viewpoints and with a temporal variation caused by object movement and lightning changes. The problem described here can be formulated as how to decide the stitching path through the overlapping region of subsequent images.
Keywords
Computer vision, field robotics, weeding.
Level
MSc student project.
Project start
Anytime
Institution
Biosystems Engineering
Institute of Chemical Engineering, Biotechnology & Environmental Technology
University of Southern Denmark
Niels Bohrs Alle 1
DK-5230 Odense M
Denmark
If you don't live in Odense, we can provide contact to the dormitory administration as well as the nearby agricultural college, that has rooms for rent on a monthly basis.