Axis enables a smarter and safer world by creating network solutions that provide insights for improving security and new ways of doing business. As the industry leader in network video, Axis offers products and services for video surveillance and analytics, access control, and audio systems. Axis has more than 3,000 dedicated employees in over 50 countries and collaborates with partners worldwide to deliver customer solutions.
Category
Automatic Control, Image Analysis, Computer Science, Probability theory
Scope
2 students completing 30 credits (20 weeks) each.
Background
Automatic 3D mapping of environments using mobile robots are done both in city environments for navigation, as well as inside for automatic retail store inventorying, etc. A drone (e.g. a quadcopter) is often used to create the maps using simultaneous localization and mapping (SLAM). This is frequently used in research projects, however in commercial applications it is still not very common due to the limited battery life of a drone, the danger of using heavy drones in crowded areas, etc. For Axis it is of interest to investigate how fixed cameras (surveillance cameras on the streets, in a store, etc.) can be used to improve the professional use of drones.
Goal
Moving complexity from the drone itself to its stationary environment; perhaps we can get the same (or better) quality of the maps with less power consumption and less weight (due to simpler hardware or less onboard processing demand) and thereby longer flight times.
The thesis could focus on one of these two topics:
Better localization feedback from fixed cameras using model knowledge
Stationary cameras can be used to estimate the pose of the drone as relative coordinates to the camera or as world coordinates given that the camera pose is known. Detections in the camera view are relatively noise with bad time resolution. Given a mathematical model of the drone (estimated using system identification) we can improve the detection accuracy and estimate intermediate states between camera frames.
Complement SLAM algorithm with fixed cameras views
If the poses and intrinsics (position, rotation, lens distortion, etc.) of the fixed cameras are not known they are hard to use for 3D-mapping. However by combining these fixed views with the views of a mobile robot we could (1) detect the pose of the fixed cameras in a relative coordinate system and (2) use the views of the fixed cameras to improve the SLAM algorithm in the mobile platform.
OK, I am interested! What do I do now?
You are valuable to us – how nice that you are interested in one of our proposals! There are a few things for you to keep in mind when applying.
23-03-2024
Ange nedan vart du önskar arbeta och glöm inte bort att ange din e-postadress!