Hyperspectral Scanning for Adaptable Insect Light Scattering Model
(2026) PHYM03 20261Combustion Physics
Department of Physics
- Abstract
- Monitoring insect populations and diversity are important to assess ecosystem health. Traditional trapping and manual identification methods are slow and labor-intensive. Entomological lidar offers an automated alternative for monitoring insects in flight, but classifying species remains a challenge. Building empirical databases through catch-and-release methods is impractical due to the enormous number of insect species. As an alternative, this thesis focuses on laying the groundwork for a physics-based model by isolating the optical properties of the insect body. Hyperspectral reflectance data of female \textit{Aedes aegypti} mosquitoes were acquired in the shortwave infrared region (960–2500 nm) across two orthogonal rotation axes (yaw... (More)
- Monitoring insect populations and diversity are important to assess ecosystem health. Traditional trapping and manual identification methods are slow and labor-intensive. Entomological lidar offers an automated alternative for monitoring insects in flight, but classifying species remains a challenge. Building empirical databases through catch-and-release methods is impractical due to the enormous number of insect species. As an alternative, this thesis focuses on laying the groundwork for a physics-based model by isolating the optical properties of the insect body. Hyperspectral reflectance data of female \textit{Aedes aegypti} mosquitoes were acquired in the shortwave infrared region (960–2500 nm) across two orthogonal rotation axes (yaw and roll) using de-polarized backscatter signal. A four-parameter diffuse reflectance model, which accounts for water and melanin absorption along with scattering parameters, was fitted to the recorded spectra, achieving more than 98% (R^2) agreement. A MATLAB framework was then developed using spherical harmonics to parameterize the angular dependence of these optical properties. This allows the framework to output the backscatter optical cross-section (OCS) as a function of pitch, yaw, and wavelength, providing foundational data for future insect lidar models. (Less)
- Popular Abstract
- Can you tell what insects are made of just by looking at them?
With hyperspectral scanning in shortwave infrared, we can extract key optical parameters (including absorption features) from any observation angles to characterize a species. In this thesis, deadly dengue fever mosquitoes are scanned to determine their directionally dependent scatter- ing properties as well as water and melanin path lengths.
Insects are the most abundant group of animals with great impact on human lives. Some are key pollinators responsible for fruitful crops, some are crucial indicators of ecosystem health, and some are vectors of deadly diseases. And yet, current methods of monitoring them (malaise traps) are labor and time consuming and invasive.... (More) - Can you tell what insects are made of just by looking at them?
With hyperspectral scanning in shortwave infrared, we can extract key optical parameters (including absorption features) from any observation angles to characterize a species. In this thesis, deadly dengue fever mosquitoes are scanned to determine their directionally dependent scatter- ing properties as well as water and melanin path lengths.
Insects are the most abundant group of animals with great impact on human lives. Some are key pollinators responsible for fruitful crops, some are crucial indicators of ecosystem health, and some are vectors of deadly diseases. And yet, current methods of monitoring them (malaise traps) are labor and time consuming and invasive. Entomological lidar emerges as an optical alternative, but reference reflectance models of known insects are needed to identify species in observations recorded in the field.
Hyperspectral scanning in the shortwave infrared region of 960-2500 nm was performed using de-polarized light on 6 female mosquitoes. The specimens were de-winged and mounted using a metal pin either on the thorax or end of abdomen, giving 3 samples per rotation axis.
Analysis of spectra from collected images has revealed melanin pathlength increased towards the head of the insects compared to the thorax and abdomen, consistent with previous studies.
While the water pathlength directionality remains inconclusive, scanning more specimens could produce a trend of increased values toward the posterior and anterior sides of the insect. The scattering parameters appear isotropic based on the collected data. When this study is repeated with larger dataset, the results can be used to inform a theoretical model of light scattering in the Aedes aegypti mosquito, a step towards a database to reference lidar observations against.
The optical parameters were extracted from the recorded reflectance spectra through fitting a model adapted from Kubelka-Munk diffuse reflectance theory of thin, poorly scattering targets. For the mosquito, this model explained over 98% of the captured spectra, the main discrepancy coming from signal in a water absorption band around 1850 nm being stronger than predicted by the model. An unexpected difficulty in the project turned out to be the light reflected from the legs of the mosquitoes. The legs were much brighter in the images than we had anticipated, acting almost like specular signal even though de-polarized light should only give us diffuse reflectance. This, together with the erratic orientation of the legs, made it difficult to automatically discern pixels from the insect from the background or the mounting pin, introducing additional irregularity into the dataset.
Finally, preliminary scanning of highly melanized insects (blue bottle fly) showed a decreased accuracy of the current reflectance model, while scanning larger targets (cricket) allowed for analysis of individual regions of the body. Through spectroscopy, we could discern higher lipid content in the abdomen and higher melanin and water content in the eye of a cricket. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9240834
- author
- Kizewska, Dorota Zofia LU
- supervisor
-
- MEng Li LU
- Mikkel Brydegaard LU
- Timotej Zuntar LU
- organization
- course
- PHYM03 20261
- year
- 2026
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- insects, scattering, lidar, hyperspectral imaging, Aedes aegypti, diffuse reflectance
- language
- English
- id
- 9240834
- date added to LUP
- 2026-06-18 12:08:32
- date last changed
- 2026-06-18 12:08:32
@misc{9240834,
abstract = {{Monitoring insect populations and diversity are important to assess ecosystem health. Traditional trapping and manual identification methods are slow and labor-intensive. Entomological lidar offers an automated alternative for monitoring insects in flight, but classifying species remains a challenge. Building empirical databases through catch-and-release methods is impractical due to the enormous number of insect species. As an alternative, this thesis focuses on laying the groundwork for a physics-based model by isolating the optical properties of the insect body. Hyperspectral reflectance data of female \textit{Aedes aegypti} mosquitoes were acquired in the shortwave infrared region (960–2500 nm) across two orthogonal rotation axes (yaw and roll) using de-polarized backscatter signal. A four-parameter diffuse reflectance model, which accounts for water and melanin absorption along with scattering parameters, was fitted to the recorded spectra, achieving more than 98% (R^2) agreement. A MATLAB framework was then developed using spherical harmonics to parameterize the angular dependence of these optical properties. This allows the framework to output the backscatter optical cross-section (OCS) as a function of pitch, yaw, and wavelength, providing foundational data for future insect lidar models.}},
author = {{Kizewska, Dorota Zofia}},
language = {{eng}},
note = {{Student Paper}},
title = {{Hyperspectral Scanning for Adaptable Insect Light Scattering Model}},
year = {{2026}},
}