Đang chuẩn bị liên kết để tải về tài liệu:
Probabilistic Event Logic for Interval-Based Event Recognition

With digital evidence, technology is always needed to process the digital data and therefore the only difference between a forensic and a non-forensic investigation of digital data is whether or not the evidence can be used in a court of law. A forensic investigation is a process that uses science and technology to develop and test theories, which can be entered into a court of law, to answer questions about events that occurred. In particular, a digital forensic investigation is a process that uses science and technology to examine digital objects and that develops and tests theories, which can be entered into a court of law,. | in Proc. IEEE Computer Vision and Pattern Recognition CVPR Colorado Springs CO 2011 Probabilistic Event Logic for Interval-Based Event Recognition William Brendel Alan Fern Sinisa Todorovic Oregon State University Corvallis OR USA brendelw@onid.orst.edu afern@eecs.oregonstate.edu sinisa@eecs.oregonstate.edu Abstract This paper is about detecting and segmenting interrelated events which occur in challenging videos with motion blur occlusions dynamic backgrounds and missing observations. We argue that holistic reasoning about time intervals of events and their temporal constraints is critical in such domains to overcome the noise inherent to low-level video representations. For this purpose our first contribution is the formulation of probabilistic event logic PEL for representing temporal constraints among events. A PEL knowledge base consists of confidence-weighted formulas from a temporal event logic and specifies a joint distribution over the occurrence time intervals of all events. Our second contribution is a MAP inference algorithm for PEL that addresses the scalability issue of reasoning about an enormous number of time intervals and their constraints in a typical video. Specifically our algorithm leverages the spanning-interval data structure for compactly representing and manipulating entire sets of time intervals without enumerating them. Our experiments on interpreting basketball videos show that PEL inference is able to jointly detect events and identify their time intervals based on noisy input from primitive-event detectors. 1. Introduction We study modeling and recognition of multiple video events that are inter-related in various ways. Such events arise in many applications including sports video where several players perform coordinated actions like running catching and passing to achieve a goal. Recognizing such events under occlusion and amidst dynamic cluttered background is challenging. We address these uncertainties by I Jointly modeling events