20/11/00 - Alan - I created this page for us to share progress, problems and thoughts!
20/11/00 - Alan - I won't be able to meet the group this afternoon as I have a job interview, so here's my progress report. I didn't get as much as I would have liked done last week as I had three coursework deadlines. I started investigating techniqu es for extracting the registration number from captured images, and got a bit more familiar with VCC. I've set aside a large chunk of time for further work this week to make up for last week, so hopefully I'll achieve something a bit more exciting.
20/11/00 - Peter - Well, like Alan I've not done as much as I'd liked to. Mainly because of deadlines, although also I've never heard anything about the Hitachi design environment being added to the machines - tutors: any word on this? I have spent a (pretty small) bit of time looking at VCC, familiarising etc. I'll be at the meeting today.
24/11/00 - Alan - I had loads of fun with VCC yesterday. It wouldn't let me create any new libraries in the group C workspace, saying "Invalid library name" for all of them. I eventually tracked it down to the name of the workspace - "SLIP group C", which caused problems because it had spaces in it - not that VCC would tell me this! I changed the workspace name to SlipC (on the middle machine of the three) and it seems to be ok now.
24/11/00 - Alan - Progress at last! I've got a camera block working that can take an image frame at will, so I can now start to move on to extracting a number plate from these images. By the way, has no-one else any news or what? Feels like I'm talking to myself..........
25/11/00 - Alan - I beefed up the site a bit, as it was, to be fair, a bit minimalist. I want to add a bit more and do a little restructuring, hopefully I'll get it done tomorrow or Monday.
26/11/00 - Morten - I have been giving some thought to the process of recognising numbers and letters. As opposed to automatically trying to find a good thresholding value I am sure that edge detection would be a whole lot easier. It is true that edge detection also requires a thresholding value, but this decides the amount of change in pixel intensities before detecting an edge (as opposed to one threshold value for the whole image). This means that edge detection will work in many different light intensities.
If our algorithm for identifying the symbols will be as Niall explained (pattern matching by subdivision of the symbols), I suggest the following process for the recognition:
26/11/00 - Morten - Added link to protocols page. I suggest that anything regarding the protocols is placed in the ~/www/protocols directory.
Added a draft for the protocol stack / partitioning of work slide to the main page of the protocols section. The .fig file resides in the ~/www/protocols directory so feel free to change anything.
Added description of LMP interface.
27/11/00 - Alan - Finally finished playing around with the web-site to make it a bit more coherant and structured. You should notice a bit more of a theme, and be able to navigate to all the main pages from any other one now. Maybe Niall and Peter could do pages for their stuff, in the same style as the others. If we document everything we are doing as we go along it's going to make it a lot easier in the end, seeing as the web-site is what we get marked on primarily; better to have to take things out than add them in.
04/12/00 - Alan - I've been working on two problems - one is the edge detection problem, and two is improving the code from the camera to let me manipulate the images more freely. I've added an image detection page which is linked off the main Applications page; the idea of this is to collate information I gather about edge detection methods, and record the outcomes of my attempts to use them.
11/12/00 - Peter - I forgot to update the news page last week, but I do now have a platform page set up. It can be found here, or is linked from the main page. I've developed most of the basic platform design, I think, but I need some input from the tutors to tell me whether I've got it right, and also how to test it to check it's doing what it ought! Also, I'll need the performance figures from Jae to complete the model.
09/01/01 - Alan - (Happy New Year!) Morten has implemented a Canny Edge Detection algorithm over Christmas, which looks like it will be good enough to use. Unfortunately, due to necessity it was developed for Linux, not VCC, so I have spent the whole day migrating it - ouch! Basically a lot of type conversions, casts and other hacks, and changing the image format used from something called SDL to bitmap. At the minute I am taking the input pictures as bitmaps outside of VCC, so an obvious target for this week is to rectify this and have them taken from within VCC; I have a fair idea of what I might have been doing wrong after consultation with Andy and Gordon Craig (thanks!), so this might not be too bad. The other items on the agenda this week are identification of numberplates from the images using the edge detection scheme, and making a start on character recognition. Further progress as it occurs.....
10/01/01 - Alan - Spent the day enhancing the edge detection stuff. It can now identify a potential numberplate from an image. It's not too sophisticated, but seems to work quite well as long as there aren't too many polygons in the source image. I have started to update the image manipulation page to bring it up to date with the stuff that has been implemented so far.
18/01/01 - Alan - As of yesterday I have been able to recognise single letters within numberplates. It's a bit tempermental though - depends on the quality of the original picture, and likes the numberplates to be exactly vertical; this is partly to do with the algorithms, but also due to the fairly low resolution of the quickcam. I have some ideas that should make the process quite a lot better though, so I hope to implement these and extend the algorithm to recognise a sequence of characters by the tutorial on Monday. You can see an abstract explanation of the method used on the appropriate section of the Application bit of the web-site.
29/01/01 - Peter - I've just updated the platform page with my latest information about the testing of the platform. So far it seems to not be going very far, though, because VCC keeps giving me errors that I don't understand. I will try to explain what I've been doing at the tutorial this afternoon.
05/02/01 - Peter - I've updated the platform page again. Not a lot of success this week, despite a lot of work in trying to solve the problems.
11/02/01 - Peter - Another update to the platform page. I've been trying to get the Bluetooth developments mapped onto my platform. As usual, there are some problems, this time with blackbox c++.
23/02/01 - Alan - Well I finally solved all the conceptual problems for migrating my code into individual whitebox C blocks. I've got all the edge detection stuff done, and Pete managed to map it to the platform ok this afternoon, HOORAY! The numberplate recognition blocks should follow pretty quickly I think.
26/02/01 - Alan - I've started working towards the final version of my web stuff, with picture and block images and descriptions of the stuff I have done so far. You can check it out on the Application page, specifically the Implementation section.