Ryan,

        Here's is my approach.
        First assumption, the weight is directly proportional to the size of the
chicken.
        With a PIC you can build a laser line scanner to draw a square grid of
lines, say  1m x 1m spaced every 10 cm or less, this grid will be draw on
the floor, if there's no chicken on the grid and the floor is flat, the
grid is perfect, if there's a chicken or more  on the grid, the image of
the grid will be a rough surface, with a CCD camera (or two for stereo
vision) you can 'see' the grid and process the image on a PC to average the
surface height and count the white spots and determine the area they occupy
inside the square, with area x height you get the volume of live meat. With
the 'live chicken density' (kg/m^3) * volume / chicken count you can get
the average weight.
        The system can be mounted on a X,Y rail or cable (driven by a PIC) for ex.
4 meters high or more depending on your roof, the idea is to move the laser
scanner and camera(s) to 'scan' the entire area.
        On the PC software you can put 'dead zones' for example where exists water
or food feeders so the scan safely skip those areas.
        There's also a possibility to scan dead chickens on the process,
during the entire scan area a map is constructed and the white spots are
marked
if a spot doesn't move say 5 scans it's assumed to be dead (or sleeping...
:) )
and can page the 'chickens administrator' to remove it. And to help the
process iluminate the local with the laser...
        Going further, with all the data collected you can trace a graphic with
food/water time/weight $ etc...
        This proccess is no intrusive so you can leave the chickens do their job
(eat,eat,eat...) without disturbing and reducing stress and territoriality
makes no diference.
        The bigger cost I see is the software on the PC and the time to develop it.
        This stuff can earn a PhD on MIT... :)








-----------------------------------
Ricardo Seixas
rseixas@pobox.com
-----------------------------------