Hyung-Tak Ju, Sae-Yong Park, Sung-Yoon Cho, Ki-Won Kwon, Tae-Ho Im, Journal of Internet Computing and Services, Vol. 26, No. 1, pp. 35-44, Feb. 2025
10.7472/jksii.2025.26.1.35, Full Text:
Keywords: explosive classification, Time series Classification, Machine Learning

Abstract

Maritime logistics, accounting for 90% of global freight volume, is a critical industry, with ports playing a key role in enhancing surveillance systems for illegal import and export goods. Among these, detecting explosives within cargo containers is a significant challenge, and X-ray systems are currently the primary method utilized. However, X-ray detection faces limitations in efficiently inspecting large containers due to constraints such as size and weight. As an alternative, research on detecting explosives using IMS (Ion Mobility Spectrometry) devices has been actively conducted. This study aims to address the limitations of existing algorithms, including KNN, TSF, and ROCKET, proposed in prior research. While KNN and TSF exhibit shorter processing times, they suffer from lower accuracy. On the other hand, the ROCKET algorithm achieves high accuracy but encounters issues with extended processing times. To overcome these challenges, this study leverages the RDST (Random Dilated Shapelet Transform) algorithm to maintain the accuracy of the ROCKET algorithm at 0.95 while reducing processing time by approximately 20 seconds, thereby improving overall performance.

Statistics
Cite this article