Abstract
The thesis aims to develop functionality to detect and monitor specific objects in the robot’s view using onboard sensors and algorithms related to mathematical image analysis/computer vision theory. Based on the current ARM powered Olympus IV development hardware platform running a Linux standard core, the computational complexity limits for realtime-like image analysis should be investigated.
The project will be in cooperation between the Mathematics department at LTH and ÅF.
Description – initial milestone
- Adding permanent WLAN communication functionality to the robot platform.
Description – image analysis milestones
Analysis of performance is done using a stationary hardware platform
- Determine suitable sensor and optics for the application at hand. Implement image analysis functionality to detect and a non-moving object using a single camera and stationary platform. Optimize and measure performance of the implemented algorithms.
- Develop image analysis functions to detect and track a moving object using a single fixed camera on a stationary platform. Optimize and measure performance of the implemented algorithms.
- Implement functionality to detect and track multiple independently moving objects using a single camera and stationary platform. Optimize and measure performance of the implemented algorithms.
- Develop an API that enables other applications running on the plattform to use the detection and tracking information.
- Add sensor for distance measurements and integrate functionlity into the existing hardware, software algorithms and APIs. Optimize and measure performance of the implemented algorithms.
Description –final milestones
- Develop a main application using the combined parts above to fulfill a
predefined use case.
- Perform tests on the fully integrated platform. Define and specify benefits and limits of using a non real-time OS (Linux) to perform realtime-like activities.
Background
Embedded devices are continuously improving their computational performance and therefore new applications of autonomous mobile platforms are constantly being devised in order to relieve humans from monotonous and hazardous tasks. This increase in problem and solution complexity results in higher requirements on accuracy of movements and requires developers to focus more on the realtime aspects of decision making and movement control. By using the popular and widespread operating system GNU/Linux, a modern, mature and powerful foundation can be created to rapidly develop potent applications and solutions to complex problems across the industry.
Student benefits
The project will give the student experience in developing software targeted for embedded GNU/Linux. Furthermore the student will gain experience implementing, using, and evaluating computer vision algorithms in real life situations.
Prerequisites
- The student is required to have good programming skills.
- The student is required to have good knowledge in mathematical image analysis.
The student must also have an interest in acquiring low-level knowledge in the fields of embedded systems and Linux.
Time table
Master’s Thesis start during spring 2008
Tutors
Ola Palm, ÅF i Lund
Pablo Cases, ÅF i Lund
Contact persons
Jörgen Malmborg, ÅF i Lund
Possible project add-ons
- Add collision detection logic
- Add collision avoidance capabilities
- Add self-localisation and navigation capabilities