Comparison of compressive imaging and video techniques for threat detection applications
High performance imaging sensors are a fundamental requirement for many defense and security applications. The usually high cost of such sensors, however, prevents their broad deployment. Modern Computational Imaging approaches like Compressed Sensing (CS) promise cost efficient sensor architectures that might enable a wider usage of some sensor technologies. However, the technological potential for military applications still has to be verified. In order to test the capabilities of a CS-system for threat detection, a software framework for automated testing was implemented. The code contains different methods for scene modulation and image reconstruction. In our previous work, we studied the classic iterative optimization methods for image reconstruction with promising, but not completely satisfactory results. Therefore, we implemented another method. This CS video method is the ‘Fourier domain regularized inversion’ (FDRI) which promises real time single pixel video imaging. In the study presented here, we compare the rather new method with the already implemented optimization approaches regarding runtime, conventional image quality metrics and suitability for threat detection applications in different spectral bands.