[postgis-devel] tester output

strk strk at keybit.net
Sat Mar 6 10:19:28 PST 2004


I've made some tests with ~ 8000 multipolgons and points
(italian municipalities boundaries and their centroids)

Here are the summaries of using the AOI/cell_area gain or
not in the estimator.

Histogram creator function acts the same for points and
polygons (no AOI/cell_area is used at that phase)

MULTIPOLYGON estm. gain = 1:
    1   (best/worst/avg)        1       1       1
    2   (best/worst/avg)        -1.001  1.651   1.228
    4   (best/worst/avg)        -1.158  41      5.456
    8   (best/worst/avg)        -1.01   205     17.659
    16  (best/worst/avg)        -1.081  214     13.073
    32  (best/worst/avg)        1       203     13.578

MULTIPOLYGON estm. gain = AOI/cell_area:
    1   (best/worst/avg)        1       1       1
    2   (best/worst/avg)        1.006   1.125   1.05
    4   (best/worst/avg)        -1.164  4.6     1.493
    8   (best/worst/avg)        -1.284  32      3.599
    16  (best/worst/avg)        -1.085  26      2.461
    32  (best/worst/avg)        1       18      2.18

POINT estm. gain = 1:
    1   (best/worst/avg)        1       1       1
    2   (best/worst/avg)        1       1.018   1.007
    4   (best/worst/avg)        1.002   1.75    1.086
    8   (best/worst/avg)        1.005   3       1.255
    16  (best/worst/avg)        1.009   5       1.323
    32  (best/worst/avg)        1.04    5       1.392

POINT estm. gain = AOI/cell_area:
    1   (best/worst/avg)        1       1       1
    2   (best/worst/avg)        1       1.002   1
    4   (best/worst/avg)        1       1.25    1.022
    8   (best/worst/avg)        1       2       1.135
    16  (best/worst/avg)        1       2       1.108
    32  (best/worst/avg)        1       3       1.123


Could you do the same kind of test  with your data ?
Note that the tester now accept a list of comma separated
bps values:
	test_estimator.pl table geom -bps 1,2,4,8,16,32 

--strk;



More information about the postgis-devel mailing list