Unsupervised learning of anomalies in function for a video surveillance system

摘要

Abstract invention patent: "unsupervised learning of anomalies in function for a video surveillance system". The present invention relates to the techniques developed to analyze a scene represented by a stream of input video frames captured by a video camera.In one method, for example, a motor learning machine can include engines of statistics to generate maps of topological function based on observations and a detection module for detecting abnormalities of function. The engines may include networks of statistical theory of re Adaptive ssonu00e2ncia (Art), whose group observed characteristics of the function of position.The statistical engine may also enhance, deteriorate, merge and delete groups. The detection module may calculate a value of a rarity on the observations and data networks recurrent in art. In addition, the sensitivity of detection can be adjusted according to the relative importance of newly observed anomalies.