Automatic Video Interpretation: A Recognition Algorithm for Temporal Scenarios Based on Pre-compiled Scenario Models

摘要

This paper presents a new scenario recognition algorithm for Video Interpretation. We represent a scenario model with the characters involved in the scenario, with its sub-scenarios and with the constraints combining the sub-scenarios. By pre-compiling the scenario models, the recognition algorithm processes temporal constraints by decomposing complex scenarios into intermediate sub-scenarios to reduce the algorithm complexity. We have tested the recognition algorithm on several videos of a bank agency to try to recognize a scenario of "Attack". We conclude by showing experimental results of the efficiency of this algorithm for real time temporal scenario recognition.