Privacy-Preserving Occupancy Estimation for Campus Buildings Using Event Cameras

Abstract

The paper reports a lightweight occupancy estimator that leverages event-camera motion signatures rather than reconstructing identifiable frames. Evaluated across three campus corridors over four weeks, the method achieved a mean absolute error of 1.4 occupants and reduced personally identifiable data exposure compared with RGB baselines. The authors include deployment guidance for institutions balancing energy optimization with responsible sensing practices.

Figure 6 compares occupancy prediction error and confidence intervals for three sensing locations under variable foot traffic.