[ Discussions | TOC | Search | Post | Reply ]
Category: Meeting Minutes
Time: 12:03:49 PM
Remote Name: 126.96.36.199
- the system to take a picture of the camera-stick setup in the creek and return a number, which will be directly related to the height of water in the creek. This number can be co-related with local rainfall.
- a confidence measure, which tells us how reliable the number mentioned above is - so as not to raise false alarms.
- the final package should be a tarred-zipped distribution, which he can download from a webpage, configure, install and run on his home computer. And it will be useful to him only if it runs on Linux.
- the log of whatever data is generated to be in simple ascii files.
- to sound the alarm send mail, beep pagers, etc.
Stephen will be putting up a 6 ft wooden log with alternating white/black tape on it. Alex wants to investigate using other colors - might be useful in contrast enhancement. Also for nighttime, a string of LEDs will be up. The setup is a video camera, at 15 min intervals the image is digitized to get a 640x480 ppm image, which at present is converted to jpg and is available on the web.
Note: For our project, we assume that the image capture setup exists - we do the image processing, the data processing.
The stick might move, but not much. For a first pass we might just ignore this; but later we should make the system robust enough to take care of this. The LEDs will reflect in still water - this can be easily taken care of.
Need to guard against system failure - LEDs lose power, etc.
A simple histogram analysis should tell us whether it is daytime/night - would be important if we need different image processing methods for day/night images. (Or we can simply check the clock)
Need to modularize the whole system, so that different image processing techniques can be plugged in at will. Initially we can use simple techniques which can later be replaced by advanced ones. Ruigang suggests we use matlab for initial prototyping.