DLP® Hyperspectral Imaging

Hyperspectral Imaging Solutions from Texas Instruments.


Hyperspectral imaging is a powerful technique for recognizing and characterizing physical materials using the principle of the varying absorption (or emission) of different wavelengths of light by different materials or objects. Hyperspectral imaging uses the techniques of spectroscopy, which usually examines homogeneous samples of material in one dimension, but expanded to two dimensions. This allows for the analysis of an entire "scene" of heterogeneous material at once. The subject can be a small field of view close-by, or can be a large remote scene. Hyperspectral imaging is used to great benefit in remote sensing of the earth from aircraft, or from orbit, to determine the characteristics of crops, forests, waterways, mineral deposits, and more.

The light used in Hyperspectral imaging may lie within the range of wavelengths visible to the human eye, or may be in the infrared or ultraviolet regions of the electromagnetic spectrum. Hyperspectral imaging requires the spreading out of light into a rainbow of wavelengths, so that the variation in light intensity versus wavelength can be measured for each point in the scene being imaged. The result is a "Hyperspectral cube", which encodes the spectrum of captured light at every point in the measured scene.

Hyperspectral imaging employs a dispersive optical element to spread the spectrum of light into spatially separated wavelengths. The dispersive element is usually a diffraction grating, which can be optimized for different spectral regions (UV, visible, NIR, etc.).

The Hyperspectral application illustrated in the diagram shows an external scene being focused by an imaging lens (like that used in an ordinary camera) on a DLP® Digital Micromirror Device (DMD). The DMD is used to decompose the image by turning on a single mirror at a time, until the entire image has been examined. The light reflected from each mirror (corresponding to each “pixel” in the scene) is collected and shined on a diffraction grating, which spreads the light out into a precisely dispersed spectrum of different wavelengths. The dispersed light is detected by a CCD (or CMOS) sensor array similar to that used in a camera, but in this case it is capturing the spectrum of each pixel, rather than capturing a broad-spectrum image of the 2 dimensional scene all at once. There is no correspondence between points in the scene, and points in the sensor array. The points in the sensor array correspond to each specific wavelength dispersed by the diffraction grating.

The embedded processor has two functions. The first is to send the commands to the DMD controller to turn on only the precise mirror at each instant to scan (or decompose) the image and send each pixel of the image to the diffraction grating. The second function of the embedded processor is to collect the data from the sensor and assemble the data into the Hyperspectral cube as the entire scene is scanned.

The embedded processor can also do further processing on the Hyperspectral cube, to display it in various views and slices, or to compare the spectral signature of the scene to stored signatures and reduce the data further – for example to identify the materials or conditions revealed by the scene.

It is interesting to note that the data in the Hyperspectral cube can be viewed as a graph of the light intensity vs. wavelength at each point in the scene. Just as with traditional spectroscopy, the distinctive shape of this curve constitutes a spectral signature of the material being examined. By comparing the spectral signature of the sample to stored reference signatures, it is possible to ascertain the physical and chemical composition of the sample.

The diagram shows a DLP® chipset, which includes the DMD, and a DMD Controller chip, plus a DMD Analog Control chip (depending on the specific DLP® chipset). Various DLP® chipsets are available, with different DMD sizes, resolutions, and other specifications. The best choice for a DLP® chipset will depend on the Hyperspectral imaging system’s specifications, such as the range of wavelengths to be measured, the resolution of the image needed, the speed of acquisition of a Hyperspectral cube, etc.

The choice of sensor array device will depend, again, on the range of wavelengths to be measured. Other considerations for the sensor include the required sensitivity, speed of acquisition, noise, temperature range, interface requirements, cost, and other factors.

The system control and signal processing is accomplished by the Embedded Processor (Such as TI OMAP®). Power is provided by TI Power devices. The details of the optical layout and components are not shown in the diagram. The diagram is intended to convey as simply as possible the overall functionality of a DLP-based Hyperspectral imaging application. An actual product will require additional optical components and optical design in order to achieve full functionality.

View more

Application Notes (2)

Title Abstract Type Size (KB) Date Views
PDF 3.85 MB 07 Oct 2013 3447
PDF 529 KB 20 Jul 2010 2109

Support and Community


Visit the TI Wiki

Other Support