Specular reflection removal for robust event-based 3D sensing

Perceiving non-Lambertian surfaces with specular reflection is challenging. Event cameras, with high dynamic range, temporal resolution, and low data redundancy, offer a promising 3D sensing tool, but specular reflections still produce dense, spatio-temporally aliased event streams that obscure surface features and hinder reliable event-based 3D sensing. This Letter presents a hardware-software co-design method for removing reflection interference. A strobed fluorescence apparatus actively modulates surface feature signals, inducing pronounced differences in event polarity and spatio-temporal distribution compared with reflection noise. Within each strobing period, events are filtered based on polarity consistency and spatio-temporal characteristics, enabling effective signal-to-noise separation. Thus, robust 3D sensing is achieved with clean encoded signals. Experiments show that the proposed method suppresses reflection interference and preserves surface details better than existing light denoising methods, achieving sensing accuracy comparable to frame-based camera systems.