Event-Based Sensing in the Underwater Environment
Abstract
We aim to investigate the applicability of neuromorphic event-based vision sensors for highspeed detection of remote and faint transWe aim to investigate the applicability of neuromorphic event-based vision sensors for highspeeddetection of remote and faint transient objects in water and through the water-air interface. The latter will be used to explore the potential to detect celestial objeient objects in water and through the water-airinterface. The latter will be used to explore the potential to detect celestial objects for localization and navigation in GPS-denied environments. Event-based vision sensors are a novel imaging device in which eachcts forlocalization and navigation in GPS-denied environments. Event-based vision sensors are anovel imaging device in which each pixel operates independently and only reports relative changes in illumination. These changes are reported asynchronously, providin pixel operates independently and only reports relativechanges in illumination. These changes are reported asynchronously, providing high temporal resolution as well as high dynamic range compared to conventional cameras. The per-pixel reporting of changes, ratheg high temporalresolution as well as high dynamic range compared to conventional cameras. The per-pixelreporting of changes, rather than periodic frame-based reporting of absolute intensities, allows these sensors to detect faint and fast-moving objects while prr than periodic frame-based reporting of absolute intensities, allowsthese sensors to detect faint and fast-moving objects while producing very little, but highly informative, data, independent of the absolute brightness level. Sampling the environment onlyat spoducing very little, but highlyinformative, data, independent of the absolute brightness level. Sampling the environment onlyat spatial locations, and at the precise time when a change occurs, results in contrast -enhanced visual imaging as relative differencesatial locations, and at the precise time when a change occurs, results in contrast -enhancedvisual imaging as relative differences are sparsely reported. Event-based vision sensors can also be combined with polarization filters to reduce environmental noise and are sparsely reported. Event-based vision sensors canalso be combined with polarization filters to reduce environmental noise and potentially improve object detection. These contrast-enhancing properties make event-based vision sensors well suited for challengi potentiallyimprove object detection. These contrast-enhancing properties make event-based visionsensors well suited for challenging environments, such as underwater, in which visibility is impaired by distance, the forward and partially polarized backscatter inng environments, such as underwater, in which visibility isimpaired by distance, the forward and partially polarized backscatter induced by objects, thevariable ambient lighting of the environment, and by particulate matter. Each source of visibility impairmentduced by objects, thevariable ambient lighting of the environment, and by particulate matter. Each source ofvisibility impairment will produce a particular pattern of spatio-temporal changes in the event driven sensory domain enabling us to actively compensate will produce a particular pattern of spatio-temporal changes in the eventdrivensensory domain enabling us to actively compensate f for these distractors and improve vision-based object detection when submerged in water.or these distractors and improvevision-based object detection when submerged in water.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 19, 2020
- Source ID
- N629092012078
Entities
People
- Gregory Cohen
Organizations
- Office of Naval Research
- United States Navy
- Western Sydney University