If the collected values of event predicates form a bi-modal
distribution, where the number of 0 and 1 values of event predicates are very
close, then an edge is impled. Based on this observation, a simple statistic and
a decision making function can be defined as the following:
151
be applied to localized edge detection for sensor networks as well. Thus, this
decision function is usually controlled by a threshold value. The larger the
then presents the key principles of two edge detection schemes extended from
approach for a sensor node would be to collect values of event predicates in its
???thicker??? will be the detected boundary.
Ren-Shiou Liu, Lifeng Sang, and Prasun Sinha
Fig. 2. If the edge is within the distance of a predefined tolerace radius for a sensor
node, S0, which is located in the event region, then, sensor S0 is claimed to be an
edge sensor.
S = 1 ??’ |p1 ??’ p0|
|p1 + p0|
, (1)
F(S) = ??1 if S ?? t0,
0 if S < t0.
(2)
In the statistical formula, that is equation (1), p0 stands for the number
of 0 values of event predicates and p1 represents the number of 1 values of
predicates. As p0 and p1 get closer, the resulting statistic value S will eventually
become larger than the pre-selected threshold t0 in the decision making
function F.
Obviously, the performance of the statistical approach heavily depends
on the choice of the threshold t0 and the probing radius R. The larger the
threshold, the higher the precision of edge detection.
Pages:
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267