HYPERSPECTRAL IMAGING SYSTEMS
Hyperspectral imaging system and methods that may be used for imaging objects in three-dimensions are disclosed. A cylindrical lens array and/or a slit array may be used to re-image and divide a field of view into multiple channels. The multiple channels are dispersed into multiple spectral signatures and observed on a two-dimensional focal plane array in real time. The entire hyperspectral data cube is collected simultaneously.
This application is a continuation of U.S. application Ser. No. 11/933,253 filed Oct. 31, 2007 which claims priority to U.S. Provisional No. 60/914,618, filed Apr. 27, 2007, and is a continuation-in-part of U.S. application Ser. No. 11/758,986, filed Jun. 6, 2007, which is a continuation of U.S. application Ser. No. 11/220,016, filed Sep. 6, 2005, which claims priority to U.S. Provisional No. 60/607,327, filed Sep. 3, 2004. U.S. application Ser. No. 11/758,986 is also a continuation-in-part of U.S. application Ser. No. 11/437,085, filed May 19, 2006, which is a divisional of U.S. application Ser. No. 10/325,129, filed Dec. 20, 2002 (now U.S. Pat. No. 7,049,597), which claims priority to U.S. Provisional No. 60/344,130, filed Dec. 21, 2001. Each of the above-referenced applications are incorporated herein by reference.
U.S. GOVERNMENT RIGHTSThe U.S. Government has certain rights in this invention as provided for by the terms of Contract #F19628-03-C-0079 awarded by the U.S. Air Force.
BACKGROUNDHyperspectral imaging is a technique used for surveillance and reconnaissance in military, geophysical and marine science applications. Objects viewed by a hyperspectral imaging system are often observed in three-dimensions, x, y (spatial) and λ (color wavelength). Spatial observations (x, y) allow a person to observe an image when high contrast is available. However, when an object is too far away to resolve, is camouflaged, or of unique chemical composition, spectral signatures help identify otherwise unobservable objects, for example to differentiate between friendly and enemy artillery.
Hyperspectral imaging typically employs a scanning slit spectrometer; although Fourier-transform imaging spectrometers (FTIS) and scanning filter (Fabry-Perot) imaging systems have also been used. These devices, however, record only two-dimensions of a three-dimensional data set at any one time. For example, the scanning slit spectrometer takes spectral information over a one-dimensional field of view (FOV) by imaging a scene onto a slit then collimating light from the slit through a dispersive element (prism) and re-imaging various wavelength images of the slit onto a detector array. In order to develop three-dimensional information, the slit is scanned over the entire scene producing different images that must be positionally matched in post-processing. The FTIS and Fabry-Perot techniques also scan: the former scans in phase space and the latter scans in frequency space.
Current scanning spectrometer designs have resulted in large, expensive and unwieldy devices that are unsuitable for hand-held or vehicle applications. While these spectrometers have been employed effectively in airborne and satellite applications, they have inherent design limitations. These limitations arise due to motion of the associated platform, motion or changes in the atmosphere, and/or motion of the objects in the image field that occur during scan sequences. Motion of the platform results in mismatched and misalignment, reducing the resolution and hence the effectiveness of the observations. At the same time, a moving object, such as a missile, may escape detection if the object is moving faster than the spectrometer scan rate.
SUMMARYIn one embodiment, a hyperspectral imaging system includes a focal plane array and a grating-free spectrometer that divides a field of view into multiple channels as bars and that reimages the bars as multiple spectral signatures onto the focal plane array.
In another embodiment, a hyperspectral imaging system includes imaging optics that form an image of an object, a focal plane array, a cylindrical lens array that forms multiple images of a pupil of the imaging optics, and a prism and grating coupled to the cylindrical lens array, to disperse the multiple images as multiple spectral signatures onto the focal plane array.
In another embodiment, a multiwavelength imager is provided. Imaging optics form an image of an object. At least one micromachined optical (MMO) element array is located at or near to an image plane of the imager, providing a spectral signature for use with a focal plane array.
In another embodiment, a hyperspectral imaging system includes imaging optics for forming an image of an object, a focal plane array for detecting spectral signatures, a slit array between the focal plane array and the imaging optic, the slit array imaging parts of the image into pupil images focused as bars, and a spectrometer for reimaging the bars as multiple spectral signatures onto the focal plane array. The imaging optics move to define which parts of the image are imaged into the pupil images.
In another embodiment, a method forms a dense hyperspectral data cube from a plurality of images. The plurality of images are sequentially captured over a period. Each of the plurality of images are sampled into multiple channels, each channel being focused into a bar. Each bar is dispersed to form a plurality of spectral signatures that are captured using a focal plane array. The captured spectral signatures are converted into a hyperspectral data cube.
A hyperspectral imaging system is disclosed herein which may achieve high instrument resolution by recording three-dimensions, two spatial dimensions (x and y) and a spectral or color dimension (λ), simultaneously. Further, the hyperspectral imager may be handheld and operate to disperse and refocus an image without using moving parts. The imaging optics may for example image faster than at most f/5.
Imaging optics 104 are illustratively shown as a Cassegrain telescope but may instead comprise optical elements (e.g., as in
Spectrometer 106 divides the image from imaging optics 104 into multiple channels, where each channel forms a pupil image that is focused as a bar 1404 in an image plane 15-15 of cylindrical lens array 1402, as shown in
As illustrated in
The images received by focal plane array 114 are captured by computer processor 116 and both the location of an image and the spectral information for that location are processed to form hyperspectral data cube 118. The data are collected in parallel and may be saved to memory and/or viewed in real time in any of the recorded wavebands. Data cubes 118 may be collected at the speed of the digital detector array, typically limited by its internal digital clock. Thus data cubes may be read, for example, at a rate between 1-1000 data cubes per second with a spectral resolution in a range of about 1-50 nm, for example.
In one example, focal plane array 114 is a CCD detector formed of 1024×1280 pixels; cylindrical lens array 1402 has 60 lenses, thereby producing 60 bars 1404; and dispersive elements 110 disperse bars 1404 into spectral signatures 1602 such that distance A′ covers 20 pixels of focal plane array 114. The CCD detector thus simultaneously captures 20 spectral bins for each of 1024×60 spatial locations.
In one embodiment, hyperspectral imaging system 1400 may be mounted upon a movable platform that allows system 1400 to scan in the Y dimension, thereby allowing the system to build a dense data-cube by interlacing the captured spectral information over time to increase resolution in the Y dimension, described in further detail below. Alternatively, a scan mirror may be used with system 1400 to build a dense data-cube by interlacing the captured spectral information over time to increase resolution in the Y dimension.
Imaging optics 104 may be omitted from the hyperspectral imager of
If cylindrical lens array 1402 is used, the field integration may be accomplished with a circular aperture of the fore optic although a rectangular aperture of the fore optic is preferred. When slit array 1702 is used, a standard aperture is used. Further, when slit array 1702 is paired with cylindrical lens array 1402 stray light suppression may occur.
The use of MMO's may reduce the overall size and complexity of the hyperspectral imaging system, as well as increase the durability of an instrument using the hyperspectral imaging system, because there are no moving parts. Since the MMO elements are micro-machined they are ideally suited for manufacturing in silicon for use in infrared imagers. Alternatively, using a low cost replicating technique, the MMO elements may be molded into epoxy on glass, for use in the visible waveband.
Multiple hyperspectral imagers may be used to cover a large field of view. For example, the exterior of a surveillance plane may be covered with multiple hyperspectral imagers. Data from the multiple imagers may be compiled into one comprehensive data set for viewing and analysis.
Alternatively, a large-scale hyperspectral imager may be fabricated according to the present instrumentalities. For example, a large-scale imager may be used in aerial or satellite applications. The costs of fabricating and transporting an imager as herein disclosed may be less than similar costs associated with a traditional hyperspectral imaging system due to the decreased number of optical components and weight thereof.
In one embodiment, illustrated in
Where an assembly wheel (e.g., assembly wheel 1300, 1350) does not include prisms (e.g., prism 2206) or other dispersion components, it may then be desirable to vary the amount of dispersion to accommodate various lens sizes and aperture (e.g., pinhole, slit) spacing. For example, dispersive element(s) 110 of systems 1400, 1700 and 1800 may be rotated to increase dispersion when large lenses or large aperture spacing is used and to decrease dispersion when small lenses or small aperture spacing is used, to sample an image. Zoom lenses may also be used beneficially with an assembly wheel having differing lens sizes and aperture spacing within the hyperspectral imaging system.
Object identification, which is more than mere recognition, may be performed by software to distinguish objects with specific spatial, spectral and temporal signatures. For example, materials from which objects in the image are made may be spectrally distinguished, e.g., in the visible range, paint on an enemy tank may be distinguished from paint on a friendly tank, while in the infrared region, a water treatment plant may be distinguished from a chemical weapons factory. The software may be trained to color code or otherwise highlight elements of the image with particular spatial and/or spectral and/or temporal signatures.
Multiple hyperspectral imagers may be used to cover a large field of view. For example, the exterior of a surveillance plane may be covered with multiple hyperspectral imagers. Data from the multiple imagers may be compiled into one comprehensive data set for viewing and analysis.
The above disclosure thus describes hyperspectral imaging systems primarily with the use of prisms for dispersion, because diffraction gratings operate over one order (and then overlap). Since the hyperspectral imaging system overlaps in space, it cannot utilize another order in the same place unless the spectral band is narrow. As in
With regard to MMO elements, various combinations of optical elements may achieve similar results, as shown in the above figures. For example, one MMO element pairs a cylindrical lens array with Fresnel prisms. But, through an array of slits, light energy is also dispersed so that each slit is spread over, e.g., twenty pixels. Thus, the MMO elements may also operate this way since each may include a micromachined array of cylinders operate like the slit array. A similar result is achieved with a slit array and Fresnel prisms (a Fresnel lens with each slit), or with a Fresnel prism array (a slit array is not needed here as each prism is, in effect, its own slit).
A large-scale hyperspectral imager may be fabricated according to the present instrumentalities. For example, a large-scale imager may be used in aerial or satellite applications. The costs of fabricating and transporting an imager as herein disclosed may be less than similar costs associated with a traditional hyperspectral imaging system due to the decreased number of optical components and weight thereof.
Certain changes may be made in the systems and methods described herein without departing from the scope hereof. It should thus be noted that the matter contained in the above description or shown in the accompanying drawings should be interpreted as illustrative and not in a limiting sense. The following claims are intended to cover all generic and specific features described herein, as well as all statements of the scope of the present method and system, which, as a matter of language, might be said to fall there between.
Claims
1. A method of forming a hyperspectral data cube, comprising the steps of:
- forming an image by imaging optics;
- dividing the formed image, by a spectrometer, into multiple channels;
- refocusing the multiple channels into multiple spectral signatures on a focal plane array;
- associating each of the multiple spectral signatures with a respective one of multiple bars formed by a corresponding one of a cylindrical lens array of the focal plane array;
- recording at least one spectral image on the focal plane array from the multiple bars; and
- processing location and spectral information of the recorded at least one spectral image to form the hyperspectral data cube.
2. The method of claim 1, further comprising the step of forming a respective pupil image from each of the multiple channels.
3. The method of claim 2, further comprising a step of focusing each respective pupil image as a corresponding one of the multiple bars on an image plane of the focal plane array.
4. The method of claim 1, further comprising a step of sampling the formed image by the cylindrical lens array.
5. The method of claim 1, wherein the step of refocusing is performed in conjunction with at least one of a collimating lens, a dispersive element, and a focusing lens.
6. The method of claim 5, wherein the multiple bars run continuously in the X-direction and are spaced in the Y-direction.
7. The method of claim 6, wherein a spatial resolution is according to the spacing of the multiple bars in the Y-direction.
8. The method of claim 6, wherein each of the multiple bars is dispersed in the Y-direction so as to not overlap with respective adjacent ones of the multiple bars.
9. The method of claim 8, wherein a direction of dispersion of the multiple bars is perpendicular to an orientation of the multiple bars.
10. The method of claim 8, wherein a length of the multiple spectral signatures is determined by a dispersive power of the dispersive element.
11. The method of claim 7, wherein the spatial resolution is further determined by at least one of a zoom collimating lens, a relay lens, and variable dispersion prism.
12. The method of claim 1, wherein the imaging optics image faster than f/5.
13. The method of claim 1, wherein the cylindrical lens array is at or near to an image plane of the imaging optics.
14. The method of claim 1, wherein the imaging optics include at least one of a Cassegrain and refractive optical elements.
15. The method of claim 1, wherein the hyperspectral data cubes is collected at a speed of a digital detector array.
16. A hyperspectral imaging system, comprising:
- a focal plane array;
- a grating-free spectrometer for dividing a field of view into multiple channels as bars and for reimaging the multiple channels as multiple spectral signatures onto the focal plane array; and
- a processor connected with the focal plane array for forming a hyperspectral data cube from the multiple spectral signatures.
Type: Application
Filed: May 7, 2012
Publication Date: Aug 30, 2012
Inventor: Andrew Bodkin (Wellesley, MA)
Application Number: 13/465,911
International Classification: G01J 3/28 (20060101);