H. Kawasaki, A. Nakamura, and Y. Kuno (Japan)
action recognition, surveillance, computer vision
We often take our time finding small objects such as a TV remote controller. This paper presents a memory aid system using computer vision to help us in such situations. The vision system consists of two modules: online and offline modules. The former keeps working all the time to detect human actions of putting an object and to record the action scene images. The latter is initiated at times to compare the recorded images and the current scene to reduce false alarms. Still, the system produces a number of false alarms because the reduction of missing errors should be the first concern. We solve this problem by showing all action images to users since humans are good at scanning images to detect necessary ones. Experimental results demonstrate that showing the images on the screen is an effective way to remind us where we put the objects.
Important Links:
Go Back
