unit: pixel
3 CONCLUSIONS Ea re... Ts 3
| window | Forstner | new
In this paper some popular location method have been LS Lal. | Haim au
analysed simply, and a new location method for straight | | 9x9 |. 0.216 | 0.039 |
line and corner has been proposed. By means of analyses pete sy due ER 1
and experements, a few conclusions have been obtained. [ibo [o890T8 [2 - 0603 o]
(1) The new method can locate accurately the straight a m UE 1
line. Its accuracy is much higher than that of Hough Lbs Lr std DU» sal
Transformation. (2) The high accuracy in the location of | 15315 | 0.140 | 0.027 |
corner point can also be acquired with the new method. a + + E
In the ideal condition, the accuracy is about 0.02 pixel E 1917 | 5i ! 0.028 |
size. (3) The new method can provide not also the b d m a ——
position of the corner point, but also the directions of two » fais is 122 2174 oe iA 825 |
lines forming the corner, which is useful to matching and md
A : : | 2131 | 0.122 | 0.023 |
recognition of image. (4) On the urban image with large L À. |
scale, the result of ihe method of least squares maíiching Table 2
may not be very good because of the influence of
projective error. But it is satisfactory to use the new
method. So the new method can be used in relative | Seat eat doutes C (nil
orientation, absolute orientation and other aspects which | (grey level) | d boy
high accuracy is requared. = | Tad
| 0 | 0.023
REFERENCES PF +=
| 2 | 0.029
Ballard, D.H. and Brown, C.M., 1982. Computer Vision. mm | remis
Fórstner, W. and Gülch, E., 1986. A Fast Operator for brem e TA nas) mi —- 1
Detection and Precise Location of Distinct Points, corners L | : 2
and Centres of Circular Feature. Processing of | 8 | 0.067
Intercommission conference of ISPRS on Fast Processing kb te meme sheet re mm M cr re 4
of Phototgrammetric Data. n 10 | 0.092
Liu, S.T. and Tsai, W.H., 1990. Moment-preserving = ——————— CELA Trotter ad
Corner Detection. Pattern Recognition, Vol. 23. use Sipe H 20)
Mikhail, E.M. and Mitchell, O.R., 1984. Detection and
sub-pixel Location of Photogrammetric Targets in Digital
Images. Photogrammetria, Vol. 39.
Trinder, J.C., 1989. Precision of digital Target Location. Ie TT om op rn op
PE&RS, Vol. 55. zilsisisgizgis| FINIS
Wong, K.W. and Wei-Hsin, H., 1986. Close-Range SAS ISS) | e od | 2 | S |
Mapping with a Solid State Camera. PE&RS, Vol. 52. a a NR PES Pe |
Tabatabai, A.J. and Mitchell, O.R., 1981. Edge Location sl elslslks Lad tts Eatin}
to Subpixel Accuracy in Didital Imagery. IEEE Trans. 5181818,3,3 ABS!
Pattern Analysis and Machine Intelligence. LUS SPA PA | 9 | = | = |
'aleigigis! --iziz
£12885) Le: 165
soe pel SL
unit: pixel élgig glgiz AH
Per pee eee SS SS Sa I SS
| window size | method 1 | method 2 | method 3 | lejeieiciec) v LN lel Si.
| 99 | V8 | 0.059 | 0.043 | Sana FESTE
bu + + eres [2 (21% eis T BIE! 8
| Hn | 0.185 | 0.03 | 0.035 | iim isivisia S ele ie
| bmp] hablo Eier
| 198 | 0.125 | 0.025 | 0.027 | [EB /2 sis | Teig!
me —t-—— loo |S = [I1
PTIT CUT IP sie
Le bali idi iur AMARE b IR d ei (ood oa
11417. |. 0.046 | 0.028 |. 0.018 | 5/3 E 5 EE |! gl xl
E-————— mp [Fie (S194 ESI 2i
| "19d9 - | 9.036 | 0.025 | pois 7! Pf | Sal Ele
- + ss Eis] gx Sis
| 2121 | 0.028 | 0.023 | 0.015 | (=i 1213 ==)"
Le | + i ) [=e Sieg ve |
*method 3: internal accuracy te e bed EL 421]
Table 1
82
