| Data Fusion
Mapping imagery, both photographic and SAR, shares common scene
structures, such as the boundaries between fields, woods, roads
and buildings. Though the brightness of the features is completely
different, the information contained in the edges can be combined
in various data fusion techniques.
Data fusion can operate on various levels:
- When one sample of information can be regarded as exact,
such as ground truth, it can be used as prior knowledge in
interpreting another data source. An example would be
supervised classification.
- High-level data fusion is concerned with combining the
processed output from each sensor. In this context we might
segment each image separately and then compare the findings.
- Low-level data fusion is concerned with combining the
original images into a joint image-interpretation process. For
example, we might segment registered SAR and optical images
jointly, followed by joint classification.
|
