Moving Object Detection for Event-based vision using Graph Spectral Clustering

Moving object detection has been a central topic of discussion in computervision for its wide range of applications like in self-driving cars, videosurveillance, security, and enforcement. Neuromorphic Vision Sensors (NVS) arebio-inspired sensors that mimic the working of the human eye. Unlikeconventional frame-based cameras, these sensors capture a stream ofasynchronous 'events' that pose multiple advantages over the former, like highdynamic range, low latency, low power consumption, and reduced motion blur.However, these advantages come at a high cost, as the event camera datatypically contains more noise and has low resolution. Moreover, as event-basedcameras can only capture the relative changes in brightness of a scene, eventdata do not contain usual visual information (like texture and color) asavailable in video data from normal cameras. So, moving object detection inevent-based cameras becomes an extremely challenging task. In this paper, wepresent an unsupervised Graph Spectral Clustering technique for Moving ObjectDetection in Event-based data (GSCEventMOD). We additionally show how theoptimum number of moving objects can be automatically determined. Experimentalcomparisons on publicly available datasets show that the proposed GSCEventMODalgorithm outperforms a number of state-of-the-art techniques by a maximummargin of 30%.