Video Segmentation, Multidimensional Time Sequence, Vector Autoregression Model
Abstract
Recently, video segmentation with time series has been recognized as an important research topic. Despite great progress in this field, existing approaches have some drawbacks. We first give an overview of existing techniques and approaches, and then we analyze the applicability of the recursive least square method, multidimensional modification of exponentially weighted stochastic approximation algorithm, methods of Kaczmarz, Shown, Brown, Chow, Trigg-Leach, Roberts-Reed, finite and infinite memory algorithms for detection of faults in multidimensional time sequences. At the end we have come to the conclusion that the Trigg-Leach method is preferable for fault detection in video time sequences. Model efficiency has been checked on video containing endoscopic operation.
