International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
classification accuracy. For corn, the masking procedure
resulted in a producer's accuracy of 77.496 and a user's
accuracy of 86.296. Pasture had a very high producer's
accuracy (10095) and a rather low user's accuracy (66.7%). It
was observed that four residue fields, one tomato field, and
two cauliflower fields were improperly included in the
pasture class. Watermelon exhibits a relatively low
producer’s accuracy of 64.3% and a user’s accuracy of 75%.
Of the 14 reference fields, 5 were omitted from this class
resulting an omission error of 35.7%. The user’s accuracy of
cauliflower (81%) is relatively high when compared to other
classes. However, this class illustrates a considerably high
error of omission (46.9%) which is caused by the exclusion
of 15 fields from this category. Both pepper and bare soil
exhibit low classification accuracy. While the producer’s and
user’s accuracies of pepper were 52.8% and 38.4%
respectively, bare soil had the lowest user’s accuracy of
23.5% and a considerably low producer’s accuracy of 57.1%.
It is evident that the spectral signatures of pepper and both
tomato and corn greately overlap in the images used in this
study.
In comparison with the all bands classification of the August
image, the multi-temporal masking classification shows
10.5% increase in overall accuracy. Clover exhibits the
highest increase of 50% in the producer’s accuracy. A
significant increase in the producer’s accuracy is also evident
for bare soil, residue, and sugar beet. However, while residue
and sugar beet exhibit 1.1% and 4.7% increase in the user’s
accuracy respectively, bare soil shows 19.4% decrease. It
appears that the number of the fields that were improperly
included in the bare soil class increased with the masking
procedure. Among the classes which do not exhibit a
significant improvement are corn, tomato, pepper, and
watermelon. A significant increase (26.7%) in the user's
accuracy of pasture indicates that some of the fields that were
improperly included in pasture in all bands classification of
the August image were correctly classified with the
sequential masking procedure. Cauliflower presents the
highest increase of 54% in the user’s accuracy. A significant
improvement in the user’s accuracy (25%) is also evident for
rice. However, both cauliflower and rice exhibit a decrease in
the producer’s accuracy. As an overall, the results show that
the masking procedure improve the accuracies and the
increase is significant for several classes.
6. CONCLUSIONS
The sequential masking classification of Landsat? ETM+
images acquired in three different dates (May, July, and
August 2000), coupled with field-based analysis, to identify
summer crops proved to be better than the uni-temporal
classifications. The overall accuracy for the all bands
classification of the May image (88.9%) was the highest. This
is attributed to the fact that the number of classes are less on
May image than that of July and August images and the
classes are spectrally distinct from each other. The overall
accuracies for the all bands classification of the July and
August images were found to be 66.8% and 70.8%
respectively. The overall accuracies for the first 4 PCs
classifications were found to be slightly lower than that of all
bands classifications.
In comparison with the uni-temporal classification results, the
sequential masking classification performs better, giving an
overall accuracy of 81.3%. The use of sequential masking
196
procedure improved the overall accuracies of the
classifications of the July and August images performed
alone by more than 10%. It appears that the masking
technique has overcome the problems caused by the spectral
overlaps between the classes. The level of classification
accuracy achieved in this study through masking procedure is
possibly high enough for crop mapping from Landsat7 ETM+
images, and better results may be achieved if additional
images are used.
Acknowledgements
The authors are grateful to State Planning Organization
(DPT) of Turkey for supporting this project.
References
Aplin, P., and Atkinson, P. M., 2001. Sub-pixel land cover
mapping for per-field classification. International Journal of
Remote Sensing, 22, pp. 2853-2858.
Aplin, P., Atkinson, P. M., and Curran, P. J., 1999. Fine
spatial resolution simulated satellite sensor imagery for land
cover mapping in the United Kingdom. Remote Sensing of
Environment, 68, pp. 206-216.
Beltran, C. M., and Belmonte, A. C., 2001. Irrigated crop
area estimation using Landsat TM imagery in La Mnacha,
Spain. Photogrammetric Engineering and Remote Sensing,
67, pp. 1177-1184.
Brisco, B., Brown, R. J., and Manore, M. J., 1989. Early
season crop discrimination with combined SAR and TM data.
Canadian Journal of Remote Sensing, 15, pp. 44-54.
Catlow, D.R., Parsell, R.J., and Wyatt, B.K., 1984. The
integrated use of digital cartographic data and remotely
sensed imagery. Earth-Orientation Applications in Space
Technology, 4, pp. 225-260.
Conese, C., and Maselli, F., 1991. Use of multi-temporal
information to improve classification performance of TM
scenes in complex terrain. ISPRS Journal of Photogrammetry
and Remote Sensing, 46, pp. 187-197.
Janssen, L.L.F., Schoenmakers, R.P.H.M., and Verwaal,
R.G., 1992. Integrated segmentation and classification of
high resolution satellite images. In: Proceedings of the
International Workshop [APR TC7: Multisource Data
Integration in Remote Sensing for Land Inventory
Applications, 7-9 September, Delft, edited by M Molenaar, L.
Janssen, and H. Van Leeuwen, pp. 65-84.
Janssen, L.L.F., M.N. Jaarsma and E.T.M. van der Linden,
1990. Integrating topographic data with remote sensing for
land-cover classification. Photogrammetric Engineering and
Remote Sensing, 56(11), pp. 1503-1506.
Kurosu, T., Fujita, M., and Chiba, K., 1997. The
identification of rice fields using multi-temporal ERS-1 C
band SAR data. International Journal of Remote Sensing. 18.
pp. 2953-2965.
Lanjeri, S., Melia, J., and Segarra, D., 2001. A multi-
temporal masking classification method for vineyard
monitoring in central Spain. /nternational Journal of Remote
Sensing, 22, pp. 3167-3186.
International
Tee
Maracci, G.,
remote sensit
Thessaloniki
Sensing, 11, 1
Mason, D.C.
Lawrence, M
digital map «
remotely-sens
Geographicai
Murakami, T
Sato. G., 2
SPOT/HRV .
Journal of Re
Panigrahy, S
rotation usin
data. /SPRS J
52, pp. 85-91
Turker, M.
detection usi
15(1), pp. 49-
