Abstract and subjects
Hyperspectral imaging sensors have proven to be powerful tools for highly selective and sensitive chemical detection applications, but have some significant operational drawbacks including a detection time-lag due to the large computational overhead of the matched filter analysis of the resulting data cubes. For applications where only a single chemical is of interest or real-time detection is desired, an intelligently designed multispectral sensor can trade high resolution and continuous spectral coverage for an in-line optical matched filter, enabling snapshot chemical detection with nearly no image processing requirements. Such a system can operate with little loss of performance, greatly reduced data volume, and at a fraction of the cost. We have recently developed a high-speed, high-resolution, programmable spectral filter based on a DLP (R) digital micro-mirror device (DMD) that mimics a conventional band-pass filter by operating on the spectrum without disturbing the underlying image. Our DMD-based filter can independently choose or reject dozens or hundreds of spectral bands and present them simultaneously to an imaging sensor, forming a complete 2D image. With this new technology, even very complicated matched filters can be implemented directly into the optical train of the sensor, producing an image highlighting the target chemical within a spectrally cluttered scene in real-time without further processing. Examples of matched-filter images recorded with our visible-spectrum prototype will be displayed, and extensions to other spectral regions will be discussed. Finally, we will discuss strategies for implementing more sophisticated clutter-suppressing matched filters on the DMD-based system, including schemes that approximate the subtlety of post-processing algorithms by utilizing the DMD's duty-cycle-based gray-scale capability.