Brown Researchers Build Imaging System That Sees Through Fog

Researchers at Brown University developed an imaging system that reconstructs and tracks moving objects obscured by dense fog or turbid water, according to a university report. The technique combines a dynamic vision sensor with a deep spiking neural network and is described in Advanced Science.

The sensor records changes in brightness at individual pixels instead of capturing complete frames. Because fog changes more slowly than a moving target, the sensor largely filters out the scattering background. The neural network processes the resulting spikes to reconstruct the object's silhouette and track its position.

In laboratory tests, reconstructed images achieved structural similarity scores of up to 96 percent. The tests used moving characters and bird silhouettes hidden by drifting fog or turbid water. The system only detects objects moving relative to the camera, loses sensitivity in very dim light, and currently produces silhouettes rather than complete grayscale images.