Method of classification of sequential high-speed flame images based on deep learning of artificial neural networks

摘要

The present invention relates to an image sequence classification method using deep learning of an artificial neural network, and the image sequence classification method for predicting the combustion state of a flame in the image processing system of the present invention is an image frame including a flame in an electrical fusion layer , receiving T sequential image frames (T is a natural number greater than or equal to 1) for a plurality of time steps, and generating corresponding feature map information. In the backbone network, in the feature map information for each time step, calculating a probability for each combustion state with respect to each other, and determining, in a later fusion layer, a combustion state of the last current time stage from the probability of each combustion state for each time stage.