Multivalued Background/Foreground Separation for Moving Object Detection

摘要: The detection of moving objects is usually approached by background subtraction, i.e. constructing and maintaining an up-to-date model the detecting as those that deviate from such a model. We adopt previously proposed approach to subtraction based on self organization through artificial neural networks, has been shown well cope with several known issues for maintenance, featuring high accuracy different types videos taken stationary cameras. Here we formulate fuzzy update procedure deal decision problems typically arising when crisp settings are involved. show experimental results higher values can be reached color video sequences represent typical situations critical object detection.