Embedded system application for sunn pest detection


Yazgaç B. G., Kırcı M.

6th International Conference on Agro-Geoinformatics, Agro-Geoinformatics 2017, Virginia, Amerika Birleşik Devletleri, 7 - 10 Ağustos 2017 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/agro-geoinformatics.2017.8047027
  • Basıldığı Şehir: Virginia
  • Basıldığı Ülke: Amerika Birleşik Devletleri
  • Anahtar Kelimeler: audio processing, Embedded system, kNN, LSF, machine learning, MFCC, sunn pest, SVM, wheat bugs
  • İstanbul Teknik Üniversitesi Adresli: Evet

Özet

© 2017 IEEE.Wheat bugs are one of the most dangerous insect types for cereal plantations. These types of insect pests have a possibility of causing 100% product loss in wheat production. This insect type consists of members from Pentatomidea and Heterotoptera. Most notorious members of this group are known as sunn pest. This type of pest can be encountered on every plantation in Eurasia. Wheat bugs prefer plantations not only for feeding but also for breeding. In the absence of control measures, wheat plantations can become overpopulated with sunn pest. Recently cultural control and biological control studies have gained attention. These control method groups are preferred over chemical control, because of healthcare reasons. Moreover, integrated control methods can be projected as the feature of the pest control, as precision agriculture applications spread. Today, beginning from the spring, plant protection experts watch for sunn pest awakening to try to avoid possible sunn pest attack. Additionally, chemical poisons are sprayed on plantations for protection reasons. These hazardous insecticides are known to be seriously damaging to human health, fauna and flora. Therefore, precision techniques for spotting these pests gained undeniable importance. Audio detection, recognition and classification methods have been used for decision making about creatures. To present day, these methods are used on insects, pests, birds, reptiles etc. successfully. In this work, a successful sound detection algorithm is applied to sound recordings of different sunn pest classes on an embedded system. A capable microcomputer is programmed to perform segmentation, feature extraction and classification procedures. Mel Frequency Cepstral Coefficients (MFCC) and Line Spectral Frequencies (LSF) methods are applied for feature extraction. Following that, different classification algorithms such as k-Nearest Neighbors (kNN) and Support Vector Machine (SVM) are applied to feature vector set. The performances of the procedures are examined in the sense of accuracy, and time consumption.