← Back to Projects
Real Time Visual Localization and Mapping of Mobile Robot in Dynamic Environment
Final Year Project
2020
Human brain is extremely powerful that it can map out the entire 3D environment. Our goal was to use a cheap single camera to help a robot localize and navigate in dynamic environments.
This project uses camera as its only sensor to build a 3D map of the room and localize itself in the built map. Dynamic objects were masked and light-invariant feature extraction was applied to handle changing scenes.
Features
Dynamic obstacle handling
Segmented dynamic objects from camera input before map updates to reduce drift and improve localization stability.
Segmented dynamic objects from camera input before map updates to reduce drift and improve localization stability.
Robust mapping in practical environments
Designed the pipeline to handle low-texture regions and lighting variation while maintaining reliable correspondence estimation.
Designed the pipeline to handle low-texture regions and lighting variation while maintaining reliable correspondence estimation.
Team Members