This study proposes a method for recognition of the behavior and number of
multiple objects without separation of the objects from images. Most conven
tional techniques of behavior recognition have used bottom-up processing, i
n which features were first extracted from images, and then the extracted f
eatures were subjected to time-series analysis. However, separation of obje
cts from images at the feature extraction stage resulted in unstable proces
sing. This study aims at stable recognition of multiobject behavior. For th
is purpose, a mechanism of selective attention is proposed. With this mecha
nism, particular image regions (focusing regions) are allotted to all state
s of the NFA. (nondeterministic finite automaton) that performs sequence an
alysis, and feature extraction (event detection) is performed inside such r
egions. This approach makes it possible to detect events irrespective of no
ise (that is, changes that may occur in the image beyond the focusing regio
ns), while nondeterministic state transition means that all possible event
sequences are analyzed; hence, the behavior of multiple objects can be reco
gnized without separation of the objects from the images. Object-specific c
olor tokens are assigned to NFA active state sets, and then are transferred
along with the state transitions, which is referred to as the object discr
imination mechanism. Introduction of this mechanism allows simultaneous mul
tiobject behavior recognition and detection of the number of objects. In ad
dition, the proposed system has been extended to treat multiview images, an
d its effectiveness has been proven experimentally. (C) 2001 Scripta Techni
ca.