Share Email Print
cover

Proceedings Paper

Crop classification using multi-temporal HJ satellite images: case study in Kashgar, Xinjiang
Author(s): Pengyu Hao; Zheng Niu; Li Wang
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

The HJ satellite constellation, characterized as high temporal resolution (4 day revisit frequency), has high potential to obtain cloud-free images covering all cruel periods for crop classification during growing season. In this paper, three HJ images (in May, July and September) were acquired, the performances of different multi-spectral HJ CCD data combinations for crop classification in Kashgar, Xinjiang were estimated using library for Support Vector Machine (LIBSVM), and ground reference data obtained in 2011 field work were used as training and validation samples. The result showed that multi-temporal HJ data has a potential to classify crops with an overall classification accuracy of 93.77%. Among the three time periods utilized in this research, the image acquired in July achieved the highest overall accuracy (86.98%) because all summer crops were under dense canopy closure. Cotton could be accurately extracted in May image (both user and produce accuracy are above 90%) because of its lower canopy closure compared with spring, the rotate crop (wheat_maize) and winter crop (wheat) at the time period. Then, the July and September combination performed as good as that of all threetime- period combination, which indicated that images obtained at cruel time periods are enough to identify crops, and the additional images improve little on classification accuracy. In addition, multi-temporal NDVI in cruel time periods of the growing season is testified efficient to classify crops with significant phenonlogical variances since they achieved similar overall accuracy to that of multi-temporal multi-spectral combination.

Paper Details

Date Published: 8 November 2014
PDF: 7 pages
Proc. SPIE 9260, Land Surface Remote Sensing II, 926005 (8 November 2014); doi: 10.1117/12.2068714
Show Author Affiliations
Pengyu Hao, Institute of Remote Sensing and Digital Earth (China)
Univ. of Chinese Academy of Sciences (China)
Zheng Niu, Institute of Remote Sensing and Digital Earth (China)
Li Wang, Institute of Remote Sensing and Digital Earth (China)


Published in SPIE Proceedings Vol. 9260:
Land Surface Remote Sensing II
Thomas J. Jackson; Jing Ming Chen; Peng Gong; Shunlin Liang, Editor(s)

© SPIE. Terms of Use
Back to Top