...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.
Nick
23 Nov 2006 10:33
| 
Comments (4)
| 
Report a problem
...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 :(
www.sunriseflat.co.uk/mobrobpres.ppt
19 Nov 2006 23:57
| Comments (0)
| Report a problem
They’re both avi files compressed with DivX, and should be fine for presentation.
Download link
19 Nov 2006 20:59
| Comments (0)
| Report a problem
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.
Anthony
19 Nov 2006 19:29
| Comments (1)
| Report a problem
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.
Nick
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
Conclusions
19 Nov 2006 12:31
| Comments (1)
| Report a problem
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.
18 Nov 2006 00:26
| Comments (0)
| Report a problem
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!
17 Nov 2006 16:51
| Comments (2)
| Report a problem
The lastest Code, CODE
Currently
ImageProcessor is still being converted but I have poked it a bit more since Tom posted it and its comming along.
17 Nov 2006 03:05
| Comments (0)
| Report a problem
Loading…
Please wait - comments are loading