Random Projections for Dimensionality Reduction of Hyperspectral Data
Event details
| Date | 06.12.2011 |
| Hour | 16:00 |
| Speaker | Dr James E. Fowler, Mississippi State University |
| Location |
ELD 220
|
| Category | Conferences - Seminars |
Principal component analysis (PCA) is often central to dimensionality reduction and compression in many applications, yet its data-dependent nature as a transform computed via expensive eigendecomposition often hinders its use in severely resource-constrained settings such as satellite-borne sensors. A process is presented that effectively shifts the computational burden of PCA from the resource-constrained sensor to a presumably more capable base-station receiver. The proposed approach, compressive-projection PCA (CPPCA), is driven by projections at the sensor onto lower-dimensional subspaces chosen at random, while the CPPCA reconstruction, given only these random projections, recovers not only the coefficients associated with the PCA transform, but also an approximation to the PCA transform basis itself. This latter approximation is driven by a novel eigenvector reconstruction based on a convex-set optimization driven by Ritz vectors within the projected subspaces. The performance of CPPCA reconstruction is considered in the specific application in which random projections effectuate spectral dimensionality reduction of hyperspectral data. In particular, the effect of such random projections on the preservation of anomalous data is investigated. The popular RX anomaly detector is derived for the case in which global anomalies are to be identified directly in the random-projection domain, and it is determined via both random simulation as well as empirical observation that strongly anomalous vectors are likely to be identifiable by the projection-domain RX detector even in low-dimensional projections. Finally, a CPPCA-based reconstruction procedure for hyperspectral imagery is developed wherein projection-domain anomaly detection is employed to partition the dataset, permitting anomaly and normal pixel classes to be reconstructed separately in order to improve the representation of the anomaly pixels.
Practical information
- General public
- Free