Compressive single-pixel imaging is a signal processing paradigm wherein high-resolution scenes are recovered from highly subsampled, randomly spatially multiplexed data acquired on a single detector. While this allows for scenes to be imaged at sub-Nyquist rates, recovery becomes an iterative process with high compute requirements. For this reason, extracting information from the randomly multiplexed measurements themselves, without reconstruction, is highly desirable. One valuable piece of information is whether a static scene experienced any significant changes, called anomalies, over the course of the measurement. In this project, the student will develop novel methods for detecting and localizing anomalies from single-pixel measurements using time-series analysis and neural networks.

Organization: Scholars