Russian |
We list the basic
concepts put in the base of build-up of the suggested positional detection scheme for objects of
arbitrary form.
1. A separate
segment of the simplest form (a smooth segment which is not containing
explicit inflections) legiblly fixes a passing contour only in the direction
of own normal. It is possible to place legiblly the contour figure only
by combination of such sliding segments.
Approximation
of contour boundaries by segments of more complicated form, which can
lead to more rigid restriction of possible object positions, returns
us to the problem of preliminary detection of the simplest segments, and,
in essence, is equivalent to the problem of object detection as a whole.
2. The technique
of of hypotheses accumulation opens in essence new engineering possibilities
of a constructive solution of the integral geometrical detection problem of contour objects
in conditions of sliding (fuzzy) initial matching of contours.
Final orientation
to the technique of direct test of hypothesis about position and orientations
of search objects, i.e. to the technique of comparison with standards
(as, for example, in [32, 33], etc.)the
for reduction of an amount of tested hypotheses forces to convert to
a problem of detection enough characteristic initial fragments of search
objects. Last problem, namely - the search problem of contour segment of
nontrivial (not simple, but composite) form principally is not easier than
the problem of detection of object as a whole.
A potential possibility
of a constructive exit from this vicious circle also is given by sliding
initial matching with subsequent solution of the positional detection problem with the hypotheses
accumulation principle.
3. Development
of the concept of contiguous segments allows to realize this possibility,
- figuratively speaking, this development holds up a determinancy
degree of initial segments (in connection with definition clearness of
the cross position and orientation) up to greatest possible level
(when the problem of detection of initial segments is still in essence
more simple than detection of the whole figure), - up to that level from
which already the technique of hypotheses accumulation can pick up
the integral identification problem costructivelly.
From these positions
the suggested positional detection scheme of arbitrary form
objects is seems to be one of the most prospective schemes of automatic
solution of this problem.
Naturally, it
is necessary to realize that this scheme has essential restriction.
It is applicable only to objects though arbitrary but basically known
- fixed - forms. This scheme is intended for work in actual conditions,
- when the detection problem can be solved only
due to rigid restrictions on models of search objects. The detection problem of objects of
nonrigid form is basically other problem, - a problem which is
solved in conditions of enough complete initial information, and, moreover,
in conditions of a possibility of its sequential purposeful improvement.
Nevertheless, here again the suggested object detection scheme can take place.
It is possible, that admissible changes of search object form will be
corresponded to regular changes of a cluster form of generated hypograms,
-similar, for example to changes light stain forms at focusing lens rocking.
Such researches would be rather useful to the conscious consideration
of geometrical distortions brought by optics, - for them maximum legible
difference from cases in essence not those objects.