Classification of Whitefly Species and Biotypes Using Deep Learning

DOI Web Site
  • SATO Hirotaka
    Graduate School of Systems and Information Engineering, University of Tsukuba
  • NAKABAYASHI Hiroki
    Graduate School of Systems and Information Engineering, University of Tsukuba
  • EBIHARA Tadashi
    Graduate School of Systems and Information Engineering, University of Tsukuba Faculty of Engineering, Information and Systems, University of Tsukuba
  • MIZUTANI Koichi
    Graduate School of Systems and Information Engineering, University of Tsukuba Faculty of Engineering, Information and Systems, University of Tsukuba
  • WAKATSUKI Naoto
    Graduate School of Systems and Information Engineering, University of Tsukuba Faculty of Engineering, Information and Systems, University of Tsukuba
  • KUBOTA Kenji
    Department of Plant Protection Research, National Agricultural and Food Research Organization

Bibliographic Information

Other Title
  • 深層学習によるコナジラミ類の微小発生音を用いた種およびバイオタイプの判別
  • シンソウ ガクシュウ ニ ヨル コナジラミルイ ノ ビショウ ハッセイオン オ モチイタ タネ オヨビ バイオタイプ ノ ハンベツ

Search this article

Abstract

Whiteflies are agricultural pests causing damage to valuable crops such as tomatoes and cucumbers, and the pesticide tolerance of whiteflies differs depending on their species and biotypes. Previously, a whitefly species and biotype identification scheme using the acoustic signatures of whiteflies was proposed, focusing on the fact that whiteflies emit a tiny acoustic signal for communication that varies depending on their species and biotypes. However, only two biotypes have been reported to have been classified so far. In this paper, we propose an advanced acoustic-based classifier to classify multiple species and biotypes [Trialeurodes vaporariorum and Bemisia tabaci (biotypes B, Q1 and Q2)] by focusing on the sound spectrogram of whiteflies. We developed a deep learning model that can classify the spectrograms of whiteflies, and we conducted experiments in an anechoic chamber. As a result, we found that the proposed classifier can classify T. vaporariorum and B. tabaci (biotypes B, Q1 and Q2) with an F-value of 96.8–100 % (mean 98.7 %) while the existing acoustic classifier can only classify them with an F value of 32.7–70.5 % (mean 60.3 %). We confirmed that the proposed classifier can classify the species and biotypes of whiteflies with almost the same accuracy as a DNA-based method.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top