Advantages Of LWIR Hyperspectral Temporal Sensing

For much of the last decade, hyperspectral imaging has been an area of active research and development and hyperspectral images have only been available to researchers. This article describes a new development by Solid State Scientific Corporation (Hollis, NH, USA) that will enable hyperspectral imaging systems to enterthe main stream of remote sensing applications.

Hyperspectral images are produced by instruments called imaging spectrometers. The development of these sensors has involved the convergence of two related but distinct technologies: spectroscopy and the remote imaging. By comparison to multispectral sensors that collect images within a few broad wavelength bands, hyperspectral sensors image data simultaneously in tens of narrow, adjacent spectral bands. These measurements make it possible to derive a nearly continuous spectrum from an element in an image. After adjustment for sensor and atmospheric are applied, these image spectra can be compared to reference spectra in order to recognize and map objects, plumes and events at long range.

SSSC designs and manufactures hyperspectral imaging systems and high-speed hyperspectral transient event measurement systems (STS) for the visible through LWIR spectral bands. In each type of sensor, object space is imaged by a camera, after dispersion by a direct vision prism (DVP). The resulting "spectral image," corresponds to a diagonal slice through a hyperspectral image cube(x, y, l), having both spatial and spectral information. Figure 1 gives the image and instantaneous spectrum of a transient event. This figure includes a schematic of the basic sensor optics and fully describes the STS system. For imaging, the signal is pre-processed through a telescope and field stop. The DVP is rotated about the optic axis, while approximately 2N frames are captured, to form a spectral image cube, with N spectral sub-bands. Using a desktop computer, the spectral image cube, principle component analysis, and other output descriptors can be obtained in less than one minute.

Using fast camera framing, the STS system can capture a spectral signature every few milliseconds from an evolving event. Simple algorithms subtract pre-event clutter from the signal to increase the signal-to-noise ratio. The spectral data extracted from each frame serves as the basis for event discrimination and identification. This allows rapid detection,track and analysis of the emissions from unresolved point objects, targets or events. Final data is output at several rates: the instantaneous spectrum and 1-pixel location of the event, location to 1/20 pixel and of the temporal evolution of the event, in several frames. Total STS processing time is a few tens of milliseconds.

The Cedip Jade camera based LWIRSTS, employs variable dispersion optics. These optics divide the spectral bandwidth of the camera into K spectral sub-bands, under computer control. The number of bands is selected to optimize target signal strength or to match particular measurement requirements.

Single band operation, K = 1, provides normal(x, y) imagery. Multiple band operation, K > 1, provides spectral(x, y, l) imagery, where spectral resolution increases with K. Capture of a new spectrum in each frame supports measurement fast transients, such as an ignition flash or an explosion. The frame rate and exposure time of the Jade camera is varied to match object or event characteristics. Figure 2 shows the LWIR-STS and a "normal" image.

Figure 3 gives LWIR-STS "images" of a scene with high signal dynamic range, a pinhole back illuminated by a hot soldering iron (>200C). The image contrast has been compressed by the display algorithm. The 14-bit digital image maintains full detail similar Figure 2. This image demonstrates the very wide "temperature" dynamic range capability of LWIR arrays, as described in the next article.

In the "spectral image" of Figure 3, the point source signal is spread along the dispersion axis of the sensor, where the spectrum is sampled by a line of detectors. More precise event location is determined line fitting these spectra as the sensor dispersion axis is rotated. This demonstrated in Figure 4.

SOURCE: FLIR Systems, Inc - OEM Cores and Components