Abstract:
Insects are so important to the natural eco system as well as to humans that they are involved
in pollination, nutrient cycling, disease spreading, medicine, aesthetics, and biological control.
Besides, monitoring insects’ activities is a crucial part of understanding their behaviour though
most of the insect monitoring techniques have their own limitations. For instance, passive
LIDAR is one of the emerging techniques that is capable of in situ monitoring of insects.
However, the detection of insect heading direction is very challenging with passive LIDAR. A
novel methodology in detecting insect flight direction for a passive LIDAR system was
developed with the use of a quadrant photodetector (Hamamatsu S4349).
A control experiment was conducted by sending beads in selected directions of the field of
view of the LIDAR system where the quadrant detector is at the focus of the telescope. The
data is sampled at a 10 kHz rate for all four channels of the quadrant detector and analysis was
done in iterations. The flight direction can easily be determined by analyzing the time domain
signal through understanding the orientation of the quadrant detector. However, this method
did not reveal whether the insects flew in transverse or longitudinal direction. In the first
iteration, insect detection was determined by considering all four channels. The next step was
selecting the data set of full width at 10% of the peak signal of the event for each channel. The
dataset was transferred into the frequency domain and then compressed with the Singular Value
Decomposition (SVD). By selecting the most dominant components after data compression,
the feature set was classified using Hierarchical Cluster Analysis (HCA). Upon investigation
of individual clusters with the use of an expert dataset, it was found that the heading direction
can be identified in both longitudinal and transverse components with the sensitivity of 96%,
specificity of 95 % and accuracy of 96 %.
