Analysis of phenological changes of high vegetation in amplitude images of SAR time series
In the task of damage detection based on SAR imagery, the handling of the textural similarity between heaps of debris surrounding damaged buildings and high vegetation poses to be a challenge. However, in previous work we showed that features exist that are sensitive to the small but distinct textural difference. Since, those analyses were based on a single data set containing only one phenological state of the vegetation, we have to analyze the stability of the textural feature used for separation. In this paper, based on one time series the variation of the SAR signature of high vegetation, especially forest areas, is studied due to phenological changes. The data are high resolution Spotlight amplitude images of TerraSAR-X, because similar data are used for damage detection. The considered textural features are statistical features of the first order as well as Haralick features. The evaluation is performed on non-overlapping patches in the forest areas and comprises the search of seasonal runs of the texture features. Additionally, the occurrence of deciduous and coniferous forests is analyzed to emphasize potential phenological differences between both species.