Investigation of Multispectral Techniques for Remotely Identifying Terrain Features and Natural Materials.

摘要

This report discusses what was learned during the five-year effort to develop multispectral techniques for remotely identifying terrain features and natural materials. During this period,multispectral data were collected at sites in four states (Pennsylvania,Kansas,Virginia,and California) and various techniques were used to machine process the data. Techniques ranged from simple amplitude gating of a single spectral band to maximum likelihood classification using ten spectral bands. All techniques provided some assistance to an interpreter,some extended specific ground observations to larger areas with acceptable accuracy,and several showed promise of remote identification of materials as an operational technique in highway site planning.