본문 바로가기
  • Home

A Study on Deep Learning-based Automatic Target Recognition System in IR Image for Intelligent Combat Management System

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(1), pp.33-40
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : November 26, 2024
  • Accepted : December 26, 2024
  • Published : January 31, 2025

Gyu-Seok Do 1 Ju-Mi Park 1 Won-Seok Jang 1 Young-Sub Yang 1 Ji-Seok Yoon 1

1한화시스템

Accredited

ABSTRACT

In this paper, we propose a system that can automatically recognize targets using infrared (IR) images from an electro-optical tracking system (EOTS) for an intelligent combat management system (CMS). Target detection through IR images is the best method in low-light environments at night for military purposes. However, IR images are relatively inferior to optical images and lack of texture information, making it difficult for CMS operators to recognize targets because there are few feature points that can be tracked. To solve this problem, this study proposes an automatic target recognition (ATR) system based on a deep learning model optimized for the IR operation environment of EOTS. First, transfer learning was employed to ensure the model’s generalization performance, and data augmentation techniques were applied to IR images to reflect elements occurring in battlefield scenarios. Additionally, model ensemble methods were utilized to enhance target recognition rates, resulting in the design of an AI model suitable for naval combat systems. As a result of quantitative analysis, the proposed method demonstrated excellent performance with an accuracy of 92% for various anti-air and anti-ship targets. Therefore, it is expected to be used as an elementary technology for intelligent CMS.

Citation status

* References for papers published after 2023 are currently being built.