Wai Ho Li and I are offering three summer research projects for Monash students in Clayton about to enter their final year. These projects cover various aspects of computer vision and robotics. They will take place alongside researchers in a shared laboratory and you will be part of weekly group discussions and presentations. Application forms and eligibility details are available here.
A fast image processing library for embedded vision systems such as smartphones
To develop a library and documentation of high performance image primitives for use in real time computer vision and augmented reality applications on embedded vision systems such as smartphones.
This project will extend a high performance image processing library by adding the following features.
- Integral image calculation (for object recognition applications)
- Arbitrary kernel Gaussian convolution (for image filtering and feature descriptors)
- Feature descriptors (ORB etc for image matching, multi-view geometry)
- Non-integer image rescaling (for multi-scale approaches)
- C/C++ software development
- Experience with SIMD architectures (SSE, MMX, NEON etc)
- Experience with Android or iOS development
Open source interface to AR Drone 2.0 unmanned air vehicle
To develop a library and documentation to interface with AR Drone 2.0 hardware components to enable custom control software including semi and fully autonomous flight.
The Parrot AR Drone 2.0 is a popular consumer Unmanned Aerial Vehicle (UAV) that can be piloted using iOS and Android smartphones or tablets. Our research group has done some preliminary reverse engineering of the Version 1.0 drone to enable custom software to run on its onboard computer (ARM) such as tracking a moving target through autonomous flight. The aim here is to develop a fully featured interface library for the Version 2.0 drone (which can do a flip on command!).
This project will create a library of software components to interface to:
- Forward facing HD camera
- Downward facing odometry camera
- Rotor control
- Inertial and magnetic sensors
- Ultrasonic altitude sensors
- Wireless communication
- Additional sensors added to UAV
- C/C++ software development
- Experience with Linux or similar OS
- Experience with hardware interfaces
Low cost 3D capture system using multiple RGB-D sensors
To develop a robust 3D capture system consisting of multiple static RGB-D sensors (Microsoft Kinect or similar) placed around an indoor space.
RGB-D sensors such as the Microsoft Kinect or Asus Xtion provide colour (RGB) and depth (D) data at much lower cost than traditional marker-based "MoCap" systems such as a VICON room (used in the movie and game industry). The goal of this project is to use multiple RGB-D sensors to produce a robust 3D capture system which will capture 3D models of moving objects in an indoor space.
The project will include the following tasks:
- Design of hardware associated with RGB-D sensor mounting (3D printing + CAD)
- Software to calibrate multiple RGB-D sensors
- Software to capture data streams from multiple RGB-D sensors in real-time
- Software to fuse data streams from multiple RGB-D sensor
- Knowledge of C++ or Java (or similar OOP language)
- Knowledge of MATLAB or similar mathematical package
- Experience with CAD (Meshlab, Solidworks etc) and libraries (OpenNI)