Construction of an advanced method for recognizing monitored objects by a convolutional neural network using a discrete wavelet transform

Vadym Slyusar,Oleksandr Chernikov,Anton Chernukha,Serhii Shevchenko,Volodymyr Timofeyev,Roman Artiukh,Stella Gornostal,Nadiia Kolpachenko,Sergey Rudakov,Mykhailo Protsenko

doi:10.15587/1729-4061.2021.238601

Abstract

The tasks that unmanned aircraft systems solve include the detection of objects and determining their state. This paper reports an analysis of image recognition methods in order to automate the specified process. Based on the analysis, an improved method for recognizing images of monitored objects by a convolutional neural network using a discrete wavelet transform has been devised. Underlying the method is the task of automating image processing in unmanned aircraft systems. The operability of the proposed method was tested using an example of processing an image (aircraft, tanks, helicopters) acquired by the optical system of an unmanned aerial vehicle. A discrete wavelet transform has been used to build a database of objects' wavelet images and train a convolutional neural network based on them. That has made it possible to improve the efficiency of recognition of monitored objects and automate a given process. The effectiveness of the improved method is achieved by preliminary decomposition and approximation of the digital image of the monitored object by a discrete wavelet transform. The stages of a given method include the construction of a database of the wavelet images of images and training a convolutional neural network. The effectiveness of recognizing the monitored objects' images by the improved method was tested on a convolutional neural network, which was trained with images of 300 monitored objects. In this case, the time to make a decision, based on the proposed method, decreased on average from 0.7 to 0.84 s compared with the artificial neural networks ResNet and ConvNets. The method could be used in the information processing systems in unmanned aerial vehicles that monitor objects; in robotic complexes for various purposes; in the video surveillance systems of important objects

Highlights

The problem of security in recent years is of key importance for the development of mankind
The most needed are systems built on the basis of machine vision and artificial intelligence, including the use of robotic and unmanned aircraft systems (UMAS) [8]
We studied the effectiveness of artificial neural networks (ANNs) for the the shape of the image of the object is distorted recognition of monitored objects in the computer environment by its own or falling shadow

Summary

Introduction

The problem of security in recent years is of key importance for the development of mankind. Resolving this issue is associated with the active evolution of monitoring systems for critical infrastructure [1]. Such facilities include large industrial enterprises, energy plants [2], chemically hazardous industries [3], and other strategic objects [4], the disrup-. The main factors threatening the safety of a monitored object (MO) include fires (explosions) [5], emissions of hazardous substances [6], radiation [7], as well as unauthorized entry of persons into the territory of MO. Devising new methods for recognizing monitored objects by artificial intelligence systems is of particular relevance

Objectives

Methods

Findings

Conclusion