Classifying remotely sensed images from urban environments is challenging. Urban land cover classes are spectrally heterogeneous and materials from different classes have similar spectral properties. Image segmentation has become a common preprocessing step that helped to overcome such problems. However, little attention has been paid to impacts of segmentation on the data's spectral information content. Here, urban hyperspectral data is spectrally classified using support vector machines (SVM). By training a SVM on pixel information and applying it to the image before segmentation and after segmentation at different levels, the classification framework is maintained and the influence of the spectral generalization during image segmentation hence directly investigated. In addition, a straightforward multi-level approach was performed, which combines information from different levels into one final map. A stratified accuracy assessment by urban structure types is applied. The classification of the unsegmented data achieves an overall accuracy of 88.7%. Accuracy of the segment-based classification is lower and decreases with increasing segment size. Highest accuracies for the different urban structure types are achieved at varying segmentation levels. The accuracy of the multi-level approach is similar to that of unsegmented data but comprises the positive effects of more homogeneous segment-based classifications at different levels in one map.
Satellite based monitoring of agricultural activities requires a very high temporal resolution, due to the highly dynamic processes on viewed surfaces. The solitary use of optical data is restricted by its dependency on weather conditions. Hence, the synergetic use of SAR and optical data has a very high potential for agricultural applications such as biomass monitoring or yield estimation.
Synthetic Aperture Radar data of the ERS-2 offer the chance of bi-weekly data acquisitions. Additionally, Landsat-5 Thematic Mapper (TM) and high-resolution optical data from the Quickbird satellite shall help to verify the derived information. The Advanced Synthetic Aperture Radar (ASAR) of the European environmental satellite (ENVISAT) enables several acquisitions per week, due to the availability of different incidence angles. Moreover, the ASAR sensor offers the possibility to acquire alternating polarization data, providing HH/HV and VV/VH images. This will help to fill time gaps and bring an additional information gain in further studies.
In the present study the temporal development of biomass from two winter wheat fields is modeled based on multitemporal and multisensoral satellite data. For this purpose comprehensive ground truth information (e.g. biomass, LAI, vegetation height) was recorded in weekly intervals for the vegetation period of 2005. A positive relationship between the normalized difference vegetation index (NDVI) of optical data and biomass could be shown. The backscatter of SAR data is negatively related to the biomass. Regression coefficients of models for biomass based on satellite data and the collected biomass vary between r<sup>2</sup>=0.49 for ERS-2 and r<sup>2</sup>=0.86 for Quickbird.
The study is a first step in the synergetic use of optical and SAR data for biomass modeling and yield estimation over agricultural sites in Central Europe.
Speckle - appearing in SAR Images as random noise - hampers image processing techniques like segmentation and classification. Several algorithms have been developed to suppress the speckle effect. One disadvantage, even with optimized speckle reduction algorithms, is a blurring of the image. This effect, which appears especially along the edges of structures, is leading to further problems in subsequent image interpretation. To prevent a loss of information, the knowledge of structures in the image could be an advantage. Therefore the proposed methodology combines common filtering techniques with results from a segmentation of optical images for an object-based speckle filtering. The performance of the adapted algorithm is compared to those of common speckle filters. The accuracy assessment is based on statistical criteria and visual interpretation of the images. The results show that the efficiency of the speckle filter algorithm can be increased while a loss of information can be reduced using the boundary during the filtering process.