RICH ring finding

1
RICH ring finding
CBM Collaboration Meeting March 9-11, 2005 at
GSI, Darmstadt

• G.Ososkov
• Laboratory of Information Technologies
• Joint Institute for Nuclear Research
• 141980 Dubna, Russia
• email ososkov_at_jinr.ru

G.Ososkov Ring finding
2
CBM RICH ring recognition
CBM RICH ring finder is demanded because the
quality of STS tracks extrapolation is, at
present, not satisfactory to provide approximate
ring centers and radii.
CBM RICH1
A fragment of the simulated CBM RICH event with
points obtained by track polongations to the RICH
photodetector plane. They are distributed almost
independently on the ring positions, so it's hard
to consider them as a ring quidance mean.
3
RICH-ring recognition algorithm
was elaborated. 1. It starts from the coarse
histogramming of source data. In Fig. 1 the
histogram for 50x50 cells is presented.
Fig. 2
Fig. 1

• 2. The next is to cluster all separate areas and
  choose all points belonging to each of those
  areas. A fragment of this histogram is shown in
  Fig.2. One can see that clustering splits this
  fragment into three areas corresponding to one or
  two overlapped rings

4
Ring centers and radii are found by 2D Hough
transfom (HT).
3. Calculate centers and radii of circles drawn
through every possible triplet of points (x i,yi
i1,2,3) from the selected group. Each time we
test distances between points and obtained radii
of a triplet to be within prescribed limits.
4. Make 2D histogram of centers obtained on the
previous step. The 100x100 histogram of xc , yc
is depicted in Fig.3. Average values of radii
for triplets from the given histogram bin are
also calculated.
Fig.3.
5
RICH-ring recognition algorithm (continue)5.
Eliminate from 2D histogram all bins with the
number of points less than given limit hmin (in
our case hmin 50).
Ring centers (above) and corresponding hits
(below)
6
RICH-ring recognition algorithm (continue)
6. As in item 2, cluster all separate areas of
xc , yc and choose all centers belonging to
each of those areas. 7. In each of these areas
approximate centers and radii according to the
following rules a. Initiate by marking all bins
as non-searched b. Look for bins with
maximum values (xmax, ymax) c. If
values in all bins adjacent to (xmax, ymax) are
not exceeding them, calculate the center as the
center of gravity and mark these bins as
searched. d. Go on until exhausting
all non-searched bins.
One of such area is shown in Fig.4 Contents of
every bin presents the number of centers
accumulated in the given bin during the
histogramming process. Two maxima are
seen corresponding to two ring centers.
Fig.4
7
An example of typical event handling is shown in
fig 5.
Fig. 5
Ring-finder status Thanks to
Boris Polishchuk the program was implemented into
CBM ROOT, but is still needed to be improved and
put in the standard view. In particular, at
present, the program is not yet tuned to
- look for both halves of rings
- resolve two overlapped rings (if
needed?) - untangle (or
eliminate) point mess produced by
d-electrons. Average time to handle one event is
less than 1 sec on 2 GHz PC.
8

Some other events
too noisy rings
small p meson rings, rmin must be adjusted
Event 2
Event 3
ellipse-like ring
9
What to do next

• 1. Complete the ringfinder to be able to work
  with data of the RICH hitproducer
• 2. Be able to handle rings cut into two halves
  between upper and the lower pats of the mirror
• 3. Create a flexible control of the ringfinder
  working parameters