Multiple Objects Detection and Tracking Using Stereo Cameras

Multiple Objects Detection and Tracking Using Stereo Cameras

Adambakkam Prakash Babu, Diwakar (TU Delft Electrical Engineering, Mathematics and Computer Science; TU Delft Computer Engineering)

van Genderen, A.J. (mentor)

Delft University of Technology

Electrical Engineering | Embedded Systems

2021

The idea of autonomous driving was merely just a dream about 5 years
ago but now, with the advancements in technology, it has become prevalent.
The aim of this thesis is to provide a low-cost approach for detecting and
tracking moving objects from a moving platform. This could be used for
an autonomous vehicle to automatically avoid moving objects. Our low cost
approach will use a raspberry-pi processor board as computation platform
and stereo cameras as sensors. The process of multiple moving objects detection is performed by initially calculating the disparity maps from the stereo
image pairs. Following this is the generation of point cloud data from the
disparity map which is followed by semantic segmentation and generation
of object proposals. An EKF based tracker is used to track the moving objects
across the frames.

Tracking
Stereo vision
stereo camera
stereo matching
open cv
Computer vision
Point Cloud
Object detection

http://resolver.tudelft.nl/uuid:7f1b5719-0687-4c9b-816f-58c135ed18d4

Student theses

master thesis

© 2021 Diwakar Adambakkam Prakash Babu