METHOD AND APPARATUS FOR COUNTING PEDESTRIANS USING ARTIFICIAL NEURAL NETWORK MODEL

摘要

An apparatus for counting pedestrians according to the present invention includes: a background region extracting unit to extract a background from an input image frame, and a background region from the input image frame based on the extracted background; a feature extracting unit to extract features from the extracted background region using a feature extracting algorithm; a main component analyzing unit to select at least one statistical feature main component by analyzing a main component among a plurality of statistical feature components representing statistical relation between the background regions and the number of the pedestrians; and an artificial neural network counting part to form a learning model based on a statistical feature main component selected among a feature number component of features extracted from a learning image frame and statistical feature components and the number of pedestrians in the learning image frame, to apply the feature number component extracted from he input image frame and the selected statistical feature main component to the learning model to count the pedestrians in the input image frame.