snicknet.net

Well, I spent 90% of today working on my school project. The basic idea is that using 2 images of an object, such as ones you would use in a panaramic photo, you can recover the 3D coordinates of the pixels allowing you to create a 3D model of the object you photographed, solely using the images + a little info on your camera. Most of this extra info is stored in the JPG files by your camera, meaning in most cases you don’t even need extra data. My project, which I took on rather blindly was to write an algorithm that did just this. I have spent most of the last 2 days trying to recover the rotation and translation made by the photographer between taking the photos. I know in principle how I am to obtain this data, but in practice I’m having much more trouble implementing it than I expected. I keep getting near 0 values for my translations and am unable to account for this. Today I made some progress on other aspects of the program, I put in some code to calculate the epipolar lines based on a single input line in one image. This will allow me to produce a bunch of matched points between the images. The problem I will end up with is a bunch of data that will generate the 3D model but not the two tiny pieces that actually make it work. I hope to consult with my professor tomorrow before or after my networks midterm on the subject. The project itself is due tomorrow evening, and I’m scared to death of making a presentation about something I was not able to do.