Master thesis and internship[BR]- Master's thesis : Study of compressive sensing in view of space imaging applications[BR]- Integration Internship

Abstract

The objective of this work is to use the compressive sensing in the field of the space exploration. The compressive sensing theory affirms that an image can be retrieved taking only fewer mea- surement with respect to the minimum number dictated by the Nyquist theory. Contrarily to the classical method of acquisition of an image, the CS technique allows to create lensless cam- eras like the Flatcam, the NoRDS-CAIC and the DiffuserCam. The drawback of this technique is the need of a decoding algorithm for the reconstruction of the original image. The reconstruction method of the images is not unique, there are classical methods, that use the total variation minimization, and the deep learning methods. This work analyzes and com- pares two classical and two deep learning methods in order to find the best method for the space application. The simulations have found that the method using the deep learning approach give optimum results. The images can be well-reconstructed already with a number of measurement that is the 30% of the size of the images in less than one second. In order to practically understand the principle of the compressive sensing, an example of the DiffuserCam has been constructed in the laboratory. The camera is composed only by a diffuser and a sensor. The experience gave great results, the images have been reconstructed with great quality in short time. Finally, the compressive sensing seems to be fascinating for the space application. This tech- nique allows to suppress the compressing board because the data are taken already compressed. The suppression of the compression board reduces the mass and especially the power budgets. Moreover, the post-processing on board allows the reduction of the downlink transmission. The compressive sensing in space exploration finds its application especially in the infrared spectral band. In fact, the infrared detectors are too expensive and the compressive sensing instruments uses the single pixel detectors that are cheaper