November 23, 2006

Could everyone put their design notes for their codiing up here

...tis for the group essay.
I’ll start the essay off if that’s ok, then if a couple of other people can expand on it after that i think that’s probably the fairest approach.

November 20, 2006


Updated slides

Ant5 Video

November 19, 2006

Check yourself before you wreck yourself

...shotgun bullets are bad for your health

Here is the first draft of the presentation… Any changes we make, i’m happy to go over in the meeting tomoro… but i’ll be cranky cos i’m gonna have to wake up before 8 :(


I’ve rendered two video’s demonstrating the stages of the system. Each video is split into 4 quadrants, showing:
  1. original video
  2. the (median dilated) blob mask
  3. the full contour image
  4. the model

They’re both avi files compressed with DivX, and should be fine for presentation.

Download link

Up todate code that works pretty well

The code: Download

So far the code displays pictures that have been filtered by the blob and then the contourProcessor applied. WIth the model overlayed in red, which is what the model percieves to be the person.(Hoi correct if im wrong)

currently all it does it display this and does not use it in a useful way, some means of calculating what direction the person is heading in needs to be created. To give the robot as well as means of grabbing the images into CImg format.

but other than that we are getting there.

Also talking to Matt in CS, said that the lists in cpp are very inefficient and it might be worth converting them back to vectors if performance is too slow.



Hi I’ve split up all the slides into the follow descriptions. Could everyone who wants to talk in the presentation please let me know and also let me know if you’d want to talk about anything specific listed below.

Introduction and project overview

Visual Person Tracking Algorithms
- overview on types of algorithms etc.

Blob overview

Histogram overview

Contour overview + Canny Operator

Comparison table – advantages/disadvantages

Our step-by-step implementation (pictorial step by step + explanation)
- Class Hierarchy and Flow Control Diagram

[1] gets image from webcam
- Normal image
- [2] sends image for processing
- choose processing type

[2a] blob-based mask
- blob picture
- when to use this feature

[2b] contour image
- contour picture
- when this feature is used

[2c] histogram-based mask
- histogram picture
- when to use this feature

[3] Contour processor (compares current and previous contour images)
-contour picture of two images in a film
- advantages of masking
- [4] Robot Moves

- Video of program running

Video of moving Robot


November 18, 2006

Filtered Contour

Almost there guys!!!


We’ve managed to start putting the classes together and neaten up some of the code. So far the colour blob, edge image, and filtering of these two work fine. The code is currently able to display this filtered image.

Histogram is broken at the moment, so Chris is going to try to fix it when he gets back from home after this weekend. The contour processor has been converted by Ant but is completely untested and isn’t implemented so that can be done tomorrow.

I’m getting some pictures done for Nick to use in the presentation and will also try to get a 4 way split movie going showing all the stages of the program. Nick’s writing the rest of the material at the moment and I’ve asked him to post the slides up when he gets half way through just so we can follow the progress.

November 17, 2006

RobotInterface.cpp attempt

RobotInterface is here

The idea is to create an instance of the RobotInterface class. When initialised it will open a dialog window showing live feed from the webcam – unfortunately the user has to select the right device (the webcam) from a combo box and click ‘initialise’ – tried all morning but couldnt find a way around this as the code given isnt abstracted away from the combo boxes etc…. My C++ skills arent up to seperating them i’m afraid!!!

Then we should be able to call RobotInterface.SaveImage(), which grabs an image and saves it to “camimage.bmp” in the app directory, which can then be loaded by the CImg stuff. Again, can’t find a way around this but as we are talking about pretty low level API calls hopefully it will be quick enough to do one or two grabs a second.

Not sure if the main file should really should be RobotInterface.h instead of .cpp – anyone help???

ps Tom – I have modified DXCapture.cpp and DXCapture.h so these files aren’t the same as the ones you will have been working on – just so you know!

Motor & Vision Resources



Ok one big Code Blob

The lastest Code, CODE


ImageProcessor is still being converted but I have poked it a bit more since Tom posted it and its comming along.

June 2023

Mo Tu We Th Fr Sa Su
May |  Today  |
         1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30      

Search this blog



Most recent comments

  • The reflective writing is submitted individually so make sure you submit it yourself tomorrow mornin… by on this entry
  • Yeh Ant's charts look great. Not sure – Ant reckons 1650 max (10% over), so if anyone has any sugges… by on this entry
  • Wow, the report looks really good :) Do you think it will matter too much if it is a couple of hundr… by on this entry
  • nice, overs better than under ;) by on this entry
  • I'll do the other bits then: introduction/overview of person tracking, testing/problems (if Ant+Hoi … by Matt on this entry

Blog archive

RSS2.0 Atom
Not signed in
Sign in

Powered by BlogBuilder