View Full Version : 2-D coordinate tracking algorithm

Terry Horn
07-26-1996, 02:37 AM
To biomch-l readers:

I am forwarding this request from a friend who is interested in tracking the
motion of microscopic rigid bodies. His image processing software allows
him to discriminate between light areas (saline medium) and dark areas (the
objects of interest) on a slide, and to estimate the location of the
centroid of each object found. However, it does NOT allow for labelling of
the individual objects, so that the digitized locations are in random,
changing order. As he has some six gigabytes of videotaped images to
analyze, "manual" frame-by-frame identification is not an option.

Does anyone know of software which, given the initial target locations and a
reasonable estimate for the maximum allowable displacement between frames,
will perform the desired identification task? I suspect that someone
involved in fluid flow research (not my area!) has written such a "best
guess" algorithm; maybe someone in a fertility clinic tracking the motion of
individual spermatozoa? The algorithm must be able to identify the
coordinate pairs corresponding to the positions of up to 50 discrete
targets, and must be robust enough to handle the discontinuities inherent
during collision.

Any suggestions would be much appreciated; I will post a summary of any
responses. Thanks in advance,

T.S. Horn, Ph.D
Research Assistant Professor
Division of Orthopaedic Surgery
UAB School of Medicine
e-mail: medt030@uabdpo.dpo.uab.edu
phone: (205) 939-6054
fax: 205-939-6057
mailing address:
TCHA Gait Analysis Center
1600 7th Ave. South
ACC Suite 418
Birmingham, AL 35233