METHOD FOR EXTRACTING FEATURE VECTOR OF VIDEO BASED ON SIMILARITY MEASUREMENT

摘要

The present invention relates to a method for extracting a feature vector for a video based on similarity measurement. The method for extracting a feature vector for a video comprises: a step (a) of receiving an experiment video and a learning video consisting of a plurality of vectors; a step (b) of setting maximum similarity which a corresponding vector can have for each vector making up the learning video, and an objective function to obtain a difference in similarity between continuous function values for the corresponding vector and other vectors, and performing neural network learning to minimize the objective function; and a step (c) of using results of neural network learning to extract a feature vector for the experiment video.;COPYRIGHT KIPO 2018