Acoustic Detection of Quarantine Insects
- Members: James Schofield, Dave Chesmore
- Start Date: August 2007
Figure 1 shows an overview of the system being developed. The first stage of the project has been focussed on developing a robust detection method using fractal dimension analysis to detect incidental acoustic signals which are produced as larvae bite through wood fibres. An estimation of the Hausdorff dimension of successive short segments of an audio recording provides measure of the complexity of each audio segment. Normalisation and thresholding is then applied to identify larvae bites which have a sufficiently differing complexity to surrounding noise.
Figure 1: System Overview
Real-time detection and classification software is being developed in Microsoft C# for use on an embedded PC platform. Real-time detection has already been successfully implemented and classification will be implemented in the coming months, initially using a combination of time domain signal coding (TDSC) and artificial neural networks (ANN). Investigation into improving TDSC, using fractal dimension analysis to aid in classification and various other classification methods is being carried out.
Figure 2: A Trichoferus Griseus bite
Figure 3: Anoplophora glabripennis adult
Image reproduced with permission from Central Science Laboratory
Figure 4 - Hylotrupes bajulus damage to wood
Image reproduced with permission from Central Science Laboratory