Akademik Veri Yönetim Sistemi
5th International Conference on Agro-Geoinformatics (Agro-Geoinformatics), Tianjin, Çin, 18 - 20 Temmuz 2016, ss.454-459
Extensive consumption of cereals as food in different domestic cousins places great demand the detection of cereal pest and struggle against them. Sunn pests such as Eurygaster integriceps, Eurygaster austriaca, Aelia rostrata and Aelia acuminata are insects with similar seasonal behaviors and dominant threat to the cereal plantations of Turkey. In this work, a microphone which works in acoustic and ultrasonic sound levels with the ability of making recordings with high frequency rate is used. Following the recording of sunn pest sounds with laboratory and outdoor conditions, the sound feature vectors are obtained with the application of different methods such as Linear Predictive Cepstral Coefficients (LPCC), Line Spectral Frequencies (LSF) and Mel Frequency Cepstral Coefficients (MFCC). By analyzing different kNN models it is shown that the automatic detection of sunn pests is possible with sound processing and machine learning methods. The best results is achieved with the overall accuracy of 93.6% using the combination of MFCC and LSF methods.