본문 바로가기
  • Home

Waste Detection Based on Dilated Faster R-CNN and Similarity Analysis Using Feature Fusion

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(4), pp.25~32
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : February 6, 2025
  • Accepted : April 22, 2025
  • Published : April 30, 2025

SiUng Kim 1 JunHyeok Go 1 JeongHyeon Park 1 Nammee Moon 1

1호서대학교

Accredited

ABSTRACT

With the increasing volume of waste generation, an effective system for efficient waste management has become essential. Waste generators must be able to attach stickers corresponding to large waste items, and collectors need to verify that the disposed waste matches the stickers before proceeding with collection. If this process is not executed smoothly, additional labor and costs may be incurred for further handling. To address this, this paper proposes a system that utilizes Dilated Faster R-CNN for accurate waste classification and detection, along with a fusion-based similarity analysis to evaluate the similarity of disposed large waste items. This system enables waste generators to classify large waste items accurately and allows collectors to verify whether the disposed waste matches the intended waste for collection, thereby enhancing the efficiency of waste disposal and collection processes. The proposed Dilated Faster R-CNN achieves a 4.01% improvement in accuracy compared to the conventional Faster R-CNN. Additionally, the fusion-based similarity analysis improves processing speed by approximately 59% compared to SIFT-based feature extraction and achieves a 3.79% improvement in accuracy over ORB-based feature extraction.

Citation status

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