Comparison of Sensor Fusion Methods for Land Cover Delineations |
| |
Abstract: | This study evaluated spaceborne radar and optical data independently and in combination for land use/cover mapping. Improved classification accuracy was obtained when these discrepant data sets were combined, often with the use of radar-derived measures such as texture. One of the three study sites had a merged sensor accuracy improvement of 18 percent over either sensor independently. Four different methods to combine the two sensor types were compared. The highest classification accuracies did not occur in all study sites with the same procedures for sensor integration. Generally, a procedure with a more equal weighting of the number of bands from each sensor was best, such as three of the Principal Components Analysis (PCA) bands from the optical data with radar texture measures. |
| |
Keywords: | |
|
|