본문 바로가기
  • Home

Damaged Golf Ball Detection through CNN-based Impact Sound Classification

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2026, 31(1), pp.227~237
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : October 1, 2025
  • Accepted : December 15, 2025
  • Published : January 30, 2026

Seokho Chung 1 Jinho Cho 1 Woonho Lee 1 Sangyul Lee ORD ID 1

1한성대학교

Accredited

ABSTRACT

In the lost golf ball market, damaged balls are often mixed into distribution, leading to quality issues. To address this, we developed a CNN-based acoustic classification system that detects damaged balls by analyzing impact sounds. A hardware device with triple-layered soundproofing enabled high-quality audio data collection. The signals were converted into Mel-Spectrograms and classified using EfficientNet-based CNN, with data augmentation enhancing generalization. Model optimization (46.9 MB→4.7 MB) enabled real-time classification in embedded environments. Field tests achieved 96.5 % accuracy, and an automated system integrating Jetson Nano and Arduino significantly improved sorting efficiency. This system enhances efficiency and product reliability in lost ball sorting, offering environmental and economic benefits.

Citation status

* References for papers published after 2024 are currently being built.