This paper describes a vision system in which the input is an image of
the robot's work area and the output is a decision that can be used t
o guide the actuator in handling an object. Position and orientation o
f the object in the work area are random. Illumination conditions are
controlled and favorable; two light sources with identical spectral ch
aracteristics are positioned to avoid shadows and highlights in the wo
rk area and to ensure a uniform mean light intensity across the quadra
nts into which the image can be divided.