Multispectral Imagery

General principles of imaging are based on the registration of electromagnetic energy in particular zones of the electromagnetic spectrum using a particular sensor. In classical photography, reflected sunlight is registered in light-sensitive film. The first photographic films were sensitive in the visible spectral zone and had only one layer to record the integral amount of acquired energy; the films (and produced photographic prints) were black-and-white. The later films were composed of three layers, sensitive to three primary colors—red, green, and blue—which allowed us to develop true-color photographs, again, in the visible spectral zone, where human eyes perceive reflected sunlight radiation. Multispectral imaging systems simultaneously obtain a series of images recorded at a number of different spectral zones, far beyond the visible light. Multispectral imaging is used in many areas of science and technology but is most widely developed in astronomy, especially in radio astronomy, and in Earth remote sensing, in particular for Land Use and Land Cover Mapping. Images taken at different wavelengths can be combined to make composite images by displaying the image for each wavelength as red, green, or blue in the final image. These composite images result in color patterns that can be used to identify surface features.

Multispectral imaging is based on principles of spectral radiometry. Radiometry is the measurement of optical radiation, which is electromagnetic radiation within the frequency range between 3 × 1,011 Hz (hertz) and 3 × 1,016 Hz. This range corresponds to wavelengths between 0.01 and 1,000 µm (micrometers) and includes the regions commonly called the ultraviolet, the visible, and the infrared. A radiometer is a device used to measure the power in electromagnetic radiation. Radiometers can use different types of detectors, for example, thermal detectors or photon detectors (photodiodes). Once the detector absorbs the energy, it quantifies it and converts it into an analog signal or, in modern radiometers, into a digital number. The analog signal can be registered in a film and produce a picture, where the image density corresponds to the amount of energy, registered by a radiometer. Modern mul-tispectral systems use digital imaging technologies, where each multispectral image consists of a two-dimensional array of picture elements (pixels) and the brightness of each pixel corresponds to the amount of registered energy for the corresponding surface element. Such multispectral images can be thought of as a multilayer data set, where each layer corresponds to a particular zone of the electromagnetic spectrum.

Multispectral imagery is widely used in remote sensing for automated image classification, which is based on differences in spectral reflectance of an object in different parts of the electromagnetic spectrum, called a spectral signature (or spectral curve). There are a fair number of different [Page 1954]techniques used to extract qualitative and quantitative information about the surface from the recorded images, using analysis of the extracted spectral signatures. Spectral differences in reflectivity at certain wavelengths, identified for the same object in different layers of a multispectral image, can be used to detect and separate objects on Earth's surface. In many cases, the spectral signature of an object can be registered by laboratory or field spectrometry, where a continuous reflectance spectrum of the sample is measured through a defined wavelength region. If such a spectral signature is known for a particular object, similar objects can be delineated, classified, and identified based on matching a discrete spectral pattern, extracted from bands of a multispectral image, against the known spectral curve of the sample object. Laboratory-based measurements of spectral signatures of many materials and objects on Earth are collected by different institutions and are available as spectral databases (spectral libraries, such as ASTER spectral library, maintained by NASA's Jet Propulsion Laboratory, or USGS digital spectral library).

...