Let's Make Robots!

Advice on BSc Final Year Project

I am currently in the final year of my Computer Science Degree, and for the dissertation as I am study Computer Vision as one of my modules I was considering a project which involved a robot which could locate and track objects. I believe though that this may be too simplistic for the course so I was wondering if anyone here could offer me some advice on the matter, as it would be very much appreciated. I'd have to site you as a source of reference as all such things have to be included in the project bibliography.  

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
XicoMBD's picture

From what I've done with computer vision (the basics: color and face tracking, line folowing, etc) I always prefered the "computer on wheels" kind of robot. Firstly I used my robot TheBox and now I'm working on/ with Magabot.

Object detection can be complex, as rogue and you said, you can try to create a database of known objects and let the robot perform different tasks according to them.

If you choose to include a computer on the robot you shouldn't have that problems and go much futher :)

As I suspect is usually the case when having to propose and then carry out a thesis, it was up to me, the student to decide what the project would be to do with. I already have some knowledge of robotics, and I have what I consider to be one of the best robots for this kind of project, the Parallax Boe-Bot, so I don't see my chances of fulfilling the proposition to be that hard.

One idea that one of the tutors gave me was to create a maze-solving robot. This I believe should be possible with this robot, as specific robots for this task are prohibitively expensive. Another idea was to create a program which was able to detect coins using a camera and determine by their size and shape which amount each one was an perhaps even give a total sum of money.

I like the idea of implementing this in a robot, but I'm unsure if its possible given the size contraints of memory and processing power available.

rogue's picture

I plan to do a lightweight robot with video processing by using a wifi webcam and a Bluetooth module.  A nearby computer connected to the wifi webcam (could just be an android phone running an app like ip webcam) can process the video, then send commands via Bluetooth serial port profile back to the robot to control movement.  It is always better to have the video processing as local as possible due to latency but I believe it could work as mentioned fairly well.

rogue's picture

You might consider implementing computer learning capability with computer vision. For example, the computer can remember what an object is called or ask the user if it does not recognise it.

Also consider that while a robot is a great idea it can set-back your progress schedule for your project if you are not confident in you electrical skills.  Not trying to say don't do it, but when it comes to a senoir project try not to bite off too much- I recommend choosing something you know you can get working in its basic state which can be added onto with a lot more functionality.  A lot of this could have to do with the claims you make about your project the first term (or not, depending on the instructor), so best not to promise too much which might be difficult to deliver.  On the other hand its probably fine to deliver more than what you "promise."

You might consider incorporating a pan and tilt kit to mount the camera and/or use an infrared or ultrasonic sensor to measure the distance of an object you want to detect/learn.  A mobile platform could be something extra you do at the end of the project if time permits.

XicoMBD's picture

Sorry

XicoMBD's picture

I didn't understand what kind of advice you want.

Can you be more specific?