Real-time imaging with a hyperspectral fovea

Abstract

We address the two dominant dilemmas encountered in attempting to demonstrate real-time hyperspectral imaging: how to record a three-dimensional spectral data cube with a conventional two-dimensional detector array and how to most efficiently transmit the spectral data cube through the information bottleneck constituted by the detector's limited space-bandwidth product. We have demonstrated a new, biologically inspired approach in which a compact hyperspectral fovea is embedded within a conventional panchromatic periphery. Combined with an intelligent scanning system this will enable hyperspectral imaging to be applied only to small regions of interest previously identified using the panchromatic periphery, thus improving the efficiency with which hyperspectral imaging can be used to recognize objects in a scene. The hyperspectral fovea is realized using a coherent optical fibre bundle that reformats a two-dimensional input image into a linear output image that acts as the input to a one-dimensional, dispersive hyperspectral imager.

Intelligent multimodal and hyperspectral sensing for real-time moving target tracking

“…In this paper, we present a novel multimodal sensor design that has both omnidirectional (panoramic) view and hyperspectral fovea, and that integrates sensing and processing for real-time target detection and signature acquisition. This design is useful for a wide range of applications in real-time target recognition [10] and surveillance [3]. This sensor platform improves or differs from previous designs [2,3,5] in literature in three aspects:…”

“…But the spectral resolution is limited to few tens of bands due to the information bottleneck constituted by the space-bandwidth product of the detector arrays. To achieve two-dimensional, video-rate, hyperspectral imaging (of hundreds of spectral bands) with megapixel spatial resolution requires images to be recorded at the order of 10 Gpixels s -1 [3]. It is very expensive to capture and to process such high-rate data in real-time.…”

“…Real-time omnidirectional (or panoramic) vision systems have been implemented by catadioptric designs using hyperbolic, parabolic or ellipsoidal mirror designs, as in [8] and [9]. Real-time panchromatic peripheral view and hyperspectral fovea have been proposed [3] to balance the requirements of both the FOV and real-time hyperspectral imaging. All of these works represent the state-of-the-art of adaptive sensor designs that have advanced the conventional sensor concepts.…”

Novel, Noninvasive Multispectral Snapshot Imaging System to Measure and Map the Distribution of Human Retinal Vessel and Tissue Hemoglobin Oxygen Saturation

“…Bulk-optic [14] and fiber-optic [15] high spectral resolution snapshot imaging systems require rearranging a two-dimensional image into a one-dimensional array before being converted to a spectral data cube by the computed tomographic imaging spectrometer. The one-dimensional image input is spectrally dispersed at the detector by the diffractive optical element.…”

A multi-object spectral imaging instrument

“…There exists a number of snapshot techniques for spectral imaging [9][10][11][12]. They have applications in astronomy and microscopy [13] and also in opthalmology [14,15].…”