Fig.26

The initial histogram of the most powerful cluster (fig.24a and fig.26b) is shown on fig.26a. Here weight of the zone is a number of hypograms marking its. Rationality of smoothing operation in this case is confirmed by a view of the initial histogram of the same cluster at size of zones reduced in one and a half time (fig.26c). Thus also the centers of zones of the space of search parameters accordingly have somewhat moved. This example shows that relatively big weight of a collateral ejection (fig.26a) is caused, first of all, by accidents of distribution of hypogram points within the limits of adjacent zones.

In [3] the operation of multiple smoothing has been suggested. In this case smoothing is done{made} iterativelly until the condition will be satisfied that at two sequential stages the same zone (besides the unique one) has the maximum sum (among all zones at the given stage). The center of this zone is accepted as a search cluster point. Though such the procedure has not shown crucial advantages in comparison with once made smoothing, it is expedient to have its in a collection of tools of the analysis of hypogram intersection. Multiple smoothing will coordinate a cluster point (zone) with its environment. If this point is on a cluster boundary it is gradually shifted in its center. The stability of this point (zone) at several sequential stages of smoothing is some performance of its common reliability. Besides, if the cluster top is not acute, i.e. consists of several zones that (due to the count of neighboring zones) one zone of maximum weight will be allocated which position will depend on a context (i.e. weightinesses of enclosing zones).

In last version of program realization of hypotheses accumulation process an attempt has been undertaken to limit an amount of the zones participating in the subsequent procedure of smoothing. For this purpose so-called areas of intersection were selected, which represent set of the zones marked more than two hypograms, and also one time marked zones directly adjoining to this set (the neighbourhood, as well as earlier, the nearest 3*3*3). However, in spite of the fact that, for example in the circumscribed experiment, execution time of the smoothing procedure on the average was reduced with 1.5 up to 0.5 s, it was additionally required about 0.5 s for detection of these areas (thus the number of analysed zones was reduced on the average in 5 times). Probably, this procedure will have essential advantages at more rigid restrictions on entrance in an area of intersection. Naturally, the operation of restriction of number of analysed zones has an essential importance at multiply repeated smoothing.