Noise Reduction And Characteristic Identification Of Wood Boring Vibration Signal By Longicorn Larvae

The control of stem-boring pests is one of the most prominent problems in forestry production.The pest species can be identified according to the vibration signals of longicorn larvae eating tree trunks.However,the vibration signal is very weak,and the signal collected from trees in the field is also easily disturbed by strong noise,which affects the accuracy of species identification.The noise reduction and feature identification of wood boring vibration by Longicorn Larvae is introduced and the contributions are made as follows:The vibration signal characteristics of longicorn larvae eating tree trunks were analyzed systematically and deeply.According to the different frequency bands of the noise signals,they are divided into three types:Baseline Wandering（BW）,White Gaussian Noise（WGN）and other noise interference（Other Noise Interference,ONI）.Furthermore,different approaches are adopted to suppress noise.A noise reduction method based on Variational Mode Decomposition（VMD）is proposed to suppress the above three types of noise.For the most influential Gaussian white noise,a noise-level-dependent sub-band soft-interval threshold noise reduction method is proposed.The vibration signal of eating tree trunk is decomposed into several sub-bands by VMD,and then the different types of noise contained in each sub-band are classified and removed.Three noise reduction effect evaluation indexes are proposed,including Root Mean Square Error（RMSE）,Signal to Noise Ratio Improvement(SNR_imp)and noise reduction efficiency（η）,which are used to analyze and evaluate the noise reduction performance of different noise reduction methods for different types of noise under different noise input levels.Wavelet Packet Transform（WPT）,Empirical Mode Decomposition（EMD）and Ensemble Empirical Mode Decomposition（EEMD）based noise reduction methods is compared with the VMD-based noise reduction method,and the advantages of the VMD-based method is demonstrated.Based on the above research results,a vibration signal analysis system for wood-boring pests is proposed.Firstly,noise reduction is performed on the signal of wood boring vibration.Then,the time domain,frequency domain and time-frequency domain features are extracted to identify pest species.Field experiments show that the sub-band soft-interval threshold noise reduction method based on VMD can suppress the interference of various types of noise effectively and the behavioral characteristics of longicorn beetle larvae boring wood were identified successfully.