(Date: 28.7.2014 – 1.8.2014)
Monday are the talks for the visit by the our new boss. Unfortunately, all the effort I put into my talk was for nothing, as we ran out of time before even coming to my part. Too bad. I’m a bit angry, but, well, that’s the way it is…
The rest of the week I code my peak finding routine for the Hough Transformation. I try out a hill climbing algorithm.
The idea is the following: After the Hough Transform for all hits, you again look at every hit individually and kind-of Hough Transform it again. But instead of saving every \((r, \alpha)\) pair, you look at the position of it in your histogram. You note down the multiplicity of the looked-at bin and move on to the next \((r, \alpha)\) pair. Again, you look at the corresponding bin’s multiplicity. If it’s higher than the last one you noted down, you replace it by the current one. If not, you keep the old one. If you do that for all \((r, \alpha)\) pairs of one hit, your climbing the hill created by the hits belonging to one track. Hence the name. You end up with the coordinates of the highest bin for every hit. Some clever combining and averaging has to be done afterwards, but you find the maxima.
This can, in theory, be done in parallel, as you can climb every hit’s hill individually. But you’re doing essentially a second Hough Transform on top (although you don’t write into the histogram, you only read, so there are no race conditions), so this adds to the execution time and lowers your algorithmic performance.
By Thursday I have a working hill climber, not in parallel though. It still has some bugs but does, what it should. And it has some, for my taste, quite tricky combining routines (with center of gravity stuff, you know).
On Thursday I have a meeting with my professor to show him my progress. We also go over the outline of my thesis and change quite a lot.
Friday noon I leave the institute for a week of vacation. Yay!