A method for controlling multiple thrusters by estimating the heading angle difference and relative distance between a target object and a navigation device using a deep neural network based on multiple camera images in a marine environment

摘要

The method for controlling a plurality of thrusters by estimating the difference in the heading angle and the relative distance between the target object and the navigation device using the deep neural network of the present invention is a method for controlling a plurality of thrusters when an image captured by the preprocessor is inputted through the camera unit, obtaining camera parameters including the number of pixels in a direction, an angle of view in a horizontal direction, and a focal length; and obtaining a plurality of three-dimensional matrices by dividing the image by the preprocessor in the horizontal and vertical directions; and generating, by a pre-processing unit, a plurality of constant values through one or more convolutional operations on the three-dimensional matrix, and aligning the obtained constant values to generate a two-dimensional informatization matrix.