University of Twente Student Theses


Mapping and monitoring oilseed rape fields using Sentinel-1 time- series data

Jiao, Yizhan (2020) Mapping and monitoring oilseed rape fields using Sentinel-1 time- series data.

Link to full-text:
(only accessible for UT students and staff)
Abstract:Rapeseed is one of the most important oilseed crops in the world, which has been used in many ways as edible oil, industrial oil, and feed. As croplands vary over space and time, accurate and updated maps of rapeseed fields are needed for studies of food security. The conventional field survey method for crop mapping is time and labor-consuming. Remotely sensed data has been proved as a credible source for crop mapping. Although optical imagery can be used to map rapeseed fields, it is highly influenced by cloud cover. This study assessed the feasibility of weather independent space-borne Synthetic Aperture Radar (SAR) Sentinel-1 data for mapping and monitoring rapeseed fields in a cloudy area of China. Backscatter coefficients (VH and VV) and decomposition features (i.e., entropy, anisotropy, angle) of Sentinel-1 time-series data were used in this study. Both Support Vector Machine and Random Forests classifiers were employed to classify rapeseed fields and their performances have been compared in terms of overall accuracy as well as kappa statistic. It was found that the highest classification accuracy was achieved by using Sentinel-1 time-series data, with overall accuracy greater than 99% and a kappa coefficient greater than 0.98 through both Random Forests and Support Vector Machine classifiers, which are significantly higher than those derived from mono-temporal Sentinel-1 data. However, it was also found that there was no statistically significant difference in classification accuracy between the use of Sentinel-1 multi-temporal data (i.e., images obtained from April and May) and Sentinel-1 time-series data. Moreover, the changing pattern and change hotspot of rapeseed fields in the Hanzhong basin between 2017 and 2019 have been successfully quantified and identified. The results from this study demonstrate that multi-temporal Sentinel-1 SAR images can be used for accurate and timely rapeseed field mapping and monitoring.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page