This paper presents the results of a study aimed at investigating the potential of Compressive Sensing (CS) technologies for optical space instruments. Besides assessing the pros and cons for a wide set of proposed instrumental concepts for space applications, the study analyzed in further detail two CS-based instrument concepts, each targeting a specific application: an UV-VIS hyperspectral imager on orbiter for stellar spectro-photometry and a MIR camera for sky observation and real-time detection of Near Earth Objects (NEO). The proposed UV-VIS hyperspectral imager relies on a classical CS approach and addresses the CS reconstruction of the full image in order to implement slitless spectrophotometry of stars. The CS-based MIR camera for NEO detection instead explores a novel approach aiming at information extraction without a prior full reconstruction of the image. Besides outlining the optical design of the instruments, its key elements and a pros and cons analysis of the architecture, this paper presents the performance assessment of these instruments for typical application scenarios by means of simulated data. The results showed that, from the point of view of data reconstruction quality, a good performance can be achieved by the designed instruments in terms of compression ratio (CR) and image reconstruction. In terms of system budgets, the CS architecture offered only some marginal benefits with respect to their traditional counterparts, mainly due to the lack of a compression board. Most advantages are instead provided in terms of downlink requirements and memory buffer.
Proc. SPIE. 10010, Advanced Topics in Optoelectronics, Microelectronics, and Nanotechnologies VIII
KEYWORDS: Reconstruction algorithms, Signal to noise ratio, Multiplexing, Image quality, Photodiodes, Sensors, Modulation, Micromirrors, Signal to noise ratio, Image processing, Digital micromirror devices
Single-pixel imaging based on multiplexing is a promising technique, especially in applications where 2D detectors or raster scanning imaging are not readily applicable. With this method, Hadamard masks are projected on a spatial light modulator to encode an incident scene and a signal is recorded at the photodiode detector for each of these masks. Ultimately, the image is reconstructed on the computer by applying the inverse transform matrix. Thus, various algorithms were optimized and several spatial light modulators already characterized for such a task. This work analyses the imaging quality of such a single-pixel arrangement, when various illumination conditions are used. More precisely, the main comparison is made between coherent and incoherent ("white light") illumination and between two multiplexing methods, namely Hadamard and Scanning. The quality of the images is assessed by calculating their SNR, using two relations. The results show better images are obtained with "white light" illumination for the first method and coherent one for the second.
The Compressive Sensing (CS) is an emergent theory that provides an alternative to Shannon/Nyquist Sampling Theorem. By CS, a sparse signal can be perfectly recovered from a number of measurements, which is significantly lower than the number of periodic samples required by Sampling Theorem. The THz radiation is nowadays of high interest due to its capability to emphasize the molecular structure of matter. In imaging applications, one of the problems is the sensing device: the THz detectors are slow and bulky and cannot be integrated in large arrays like the CCD. The CS can provide an efficient solution for THz imaging. This solution is the single pixel camera with CS, a concept developed at Rice University that has materialized in several laboratory models and an IR camera released on the market in 2013. We reconsidered this concept in view of THz application and, at present, we have an experimental model for a THz camera. The paper has an extended section dedicated to the CS theory and single pixel camera architecture. In the end, we briefly presents the hardware and software solutions of our model, some characteristics and a first image obtained in visible domain.
Although nowadays spectrometers reached a high level of performance, output signals are often weak and traditional slit spectrometers still confronts the problem of poor optical throughput, minimizing their efficiency in low light setup conditions. In order to overcome these issues, Hadamard Spectroscopy (HS) was implemented in a conventional Ebert Fastie type of spectrometer setup, by substituting the exit slit with a digital micro-mirror device (DMD) who acts like a coded aperture. The theory behind HS and the functionality of the DMD are presented. The improvements brought using HS are enlightened by means of a spectrometric experiment and higher SNR spectrum is acquired. Comparative experiments were conducted in order to emphasize the SNR differences between HS and scanning slit method. Results provide a SNR gain of 3.35 favoring HS. One can conclude the HS method effectiveness to be a great asset for low light spectrometric experiments.
The paper presents a method for the simultaneous analysis of a collection of satellite images derived from different sources. The images are multispectral, multisensor, multitemporal and synthetically generated images. All these images must have the same dimension, the same resolution, and they must refer to the same geographical area. The images are organized in a parallel structure that form a 3-D block of data. We analyze this 3-D block of data using the 3- D sliding window Fourier transform (SWFT) applied on volumes of size 8 X 8 X 8. The reasons for using this strategy are: (1) the SWFT is a technique which leads to good results in 1-D signals processing like vocal signals. (2) Measurements of the receptive fields of simple cells in visual cortex having shown them to be like Gaussian modulated sinusoids. (3) The transform on the third dimension does the fusion of the different types of data included in the original multimodal image. After the computation of the 3-D transformed images we used a clustering procedure in order to reduce the dimensionality of the transformed data. To achieve a great flexibility in the selection of the significant images a slightly modified k means algorithm was used.