ACOUSTIC PEST DETECTION AND NOISE REDUCTION TECHNIQUES FOR PRECISION AGRICULTURE

Abstract

In this paper, we have provided a noninvasive acoustic sensing system that can be employed for detecting agricultural pests through the analysis of their bio-acoustic signals. The proposed system focuses on Tribolium confusum, also known as the confused flour beetle. For collecting the bio-acoustic signals generated by the insects, sensitive microphones were used while maintaining controlled experimental conditions. The collected signals were then filtered in MATLAB using FIR and IIR band pass filters in order to reduce environmental noise from wind and agricultural machinery. The next step was extracting spectral information from the signals and classifying them into respective categories employing deep learning models such as Convolutional Neural Network (CNN) and Support Vector Machine (SVM). Experiments confirmed that the background noise can be efficiently reduced using the described algorithm with almost 10 dB improvement in signal intensity.