Stored Grain Insect Detection Based On Acoustic Signal

The storage pests mostly live in the interior of the granary,and the traditional manual testing methods cannot meet the requirements of grain storage in the current situation.In order to detect and select appropriate control measures accurately and quickly,the technique of detecting acoustic signals of pests in stored grain is discussed in this thesis.The main results are as following:A novel stored grain insect acoustic signal processing method is proposed.First,the acoustic signal is denoised using wavelet soft threshold filtering technique.Next,the dual-tree complex wavelet transform is employed to analyze the de-noised signal.As a result,the duration and characteristic frequency of the pest eating sound are extracted.The experimental results show that the proposed method can detect stored-insects effectively.In a simple anechoic chamber,the sound of three stored grain pests(i.e.Sitophilus zeamais,Trogoderma variabile and Cryptolestes ferrugineus)live in wheat and corn is analyzed.Furthermore,the time-domain and frequency-domain characteristics of acoustic signals from these three pest species were extracted.The insect eating sound characteristics of three pests live in the single grain are significant different.The acoustic frequency of Sitophilus zeamais living in wheat mostly focused on 4500～5200 Hz.While the frequency band of Cryptolestes ferrugineus is 7500～8200 Hz,and Trogoderma variabile is 4200～5000Hz.Of the eating duration of three pests,Trogoderma variabile living in wheat is the longest,Sitophilus zeamais is shorter and Cryptolestes ferrugineus is the shortest.The eating sound characteristics of the same species of pests in different grains are also different.The acoustic frequency of Sitophilus zeamais living in corn is in the range of10500～12000Hz,which is obviously different from the species living in wheat.A pest detection method is proposed to identify the specie of pests using the eating sound characteristics.While the occurrence density of pests is detected by the number of eating action in unit time.Experiment results show that the accuracy is above 90% in the species and occurrence densities detection of stored grain pests,and the detecting limit of pest number exceeds one pest inper kilogram grain.A pest detection experiment was carried out in the field of grain depot.The occurrence density of Cryptolestes ferrugineus was detected successfully,and the accuracy rate reaches 90.6%.The results provide the foundation for stored grain pests automatic detecting and appropriate control measures selecting.