Hi trackers
I would like to commit changes to BrTrackFollowFinder.
Major changes :
***Use BrLocalTrack::Fit to determine the search direction when there is
3+ hits in the track. For 2 hits the BrLocalTrack::Fit for some reason
don't do as well as just drawing a line. In general however it does better.
***After each loop through the rows the tracks are grouped together and
the worst tracks are removed. First then are the hits marked as used.
Before you would risk marking a lot of hits as used for the next rows with
a large search window, even though no track was found.
***Added new tracking method {SetTrackingMethod ( 1 ) }. The only
difference between the old ( 0 ) and the new is that for the new, the
search window for the first hit covers all of the front, and the search
windows for 2+ hit tracks have a constant angle rather than a constant
window size. This improves the chance of finding the tilted tracks.
Minor changes :
*For some reason I don't know, a few times the cluster width is set to 0.
In BrLocalTrack::Fit I have added a check to see if dx or dy is 0. They
are then set to 0.02 and the method Warning is called.
More detailed notes and nice pictures can be found on.
http://alf.nbi.dk/~pchristi/BRAHMS/TRACKING/update1.html
The effect of the changes have been checked visually and seems to find
most tracks. The worst problem seems to be with tracks where there is
cluster(s) missing because of a low energysignal and if 2 tracks are so
close that the clusters are not sperated/deconvoluted.
If I hear no objections I will commit the changes tomorrow.
Tracking stuff that I will be thinking about next :
Efficiency tester
Clustering/deconvolution
Getting tracks where there are missing clusters
Ciao
Peter
:-) --------------------------------- )-:
|Peter H L Christiansen aka PAN @ NBI |
|EMAIL : pchristi@nbi.dk |
|OFFICE : Tb1 @ NBI |
|PHONE : 353 25269 <- New!! |
|SNAIL : Sdr. Fasanvej 14 ST 2000 F |
|PHONE : 38 872042 |
:-D --------------------------------- \-:
This archive was generated by hypermail 2b29 : Tue May 23 2000 - 12:01:08 EDT