Browsing by Subject "Photogrammetry"

The explosion in the popularity and success of deep learning in the past decade has accelerated the development of computationally efficient, GPU-accelerated frameworks, such as TensorFlow and PyTorch, for rapid prototyping of neural networks. In this dissertation, we show that these deep learning tools are also well-suited for computational 3D imaging problems, specifically optical diffraction tomography (ODT), photogrammetry, and our newly proposed optical coherence refraction tomography (OCRT). Underlying these computational 3D imaging techniques is a physical model that demands multiple measurements taken with either angular diversity, wavelength diversity, or both. This requirement can be compactly summarized as wavevector (or k-vector) diversity, where the magnitude and direction of the wavevector correspond to the color and angle of the light, respectively.

To understand the importance of wavevector diversity for 3D imaging, this dissertation starts by advancing a unified k-space theory of optical coherence tomography (OCT), the most comprehensive and inclusive theoretical description of OCT to date that not only describes the transfer functions of all major forms of OCT and other coherent techniques (e.g., confocal microscopy, holography, ODT), but also includes the fundamental concepts of OCT, such as speckle, dispersion, aberration, and the tradeoff between lateral resolution and depth of focus (DOF).

Consistent with this unified theory, we implemented in TensorFlow a reconstruction algorithm for ODT, a technique that relies on illumination angular diversity to achieve 3D refractive index (RI) imaging. We propose a new method for filling the well-known “missing cone” of the ODT transfer function by reparameterizing the 3D sample as the output of an untrained neural network known as a deep image prior (DIP), which we show to outperform traditional regularization strategies.

Next, we introduce OCRT, a computational extension of OCT that incorporates extreme angular diversity over OCT's already high wavelength diversity to enable resolution-enhanced, speckle-reduced reconstructions that overcome the lateral-resolution-DOF tradeoff. OCRT also jointly reconstructs quantitative RI maps of the sample using a ray-based physical model implemented in TensorFlow. We also demonstrate spectroscopic OCRT (SOCRT), an extension of spectroscopic OCT (SOCT) that overcomes its tradeoff between spectral and axial resolution.

Motivated to make OCRT more widely applicable, we propose a new use of conic-section (e.g., parabolic, ellipsoidal) mirrors to allow fast multi-view imaging over very high angular ranges (up to 360°) using galvanometers without requiring sample rotation. We theoretically characterize the achievable fields of view (FOVs) as a function of many imaging system parameters (e.g., NA, wavelength, incidence angle, focal length, and telecentricity). Based on these predictions, we constructed a parabolic-mirror-based imaging system that facilitates multi-view OCT volume capture with millimetric FOVs over up to ±75°, which we combined to perform 3D OCRT reconstructions of zebrafish, fruitfly, and mouse tissue.

Finally, we adapted the OCRT reconstruction algorithm to photogrammetric 3D mesoscopic imaging with tens-of-micron accuracy, using a sequence of smartphone camera images taken at close range under freehand motion. 3D estimation was possible due to the angular diversity afforded by the nontelecentricity of smartphone cameras, using a similar ray-based model as for OCRT. We show that careful modeling of lens distortion and incorporation of a DIP are both pivotal for obtaining high 3D accuracy using devices not designed for close-range imaging.

Environmental factors are a crucial determinant of an animals fitness. The effects of environment on fitness are often mediated by behavioral mechanisms as well as mechanisms that are ‘under the skin,’ such as growth and physiology. In my dissertation work, I study how two environmental factors – dominance rank and early-life conditions – are associated with growth and physiology. My colleagues and I test these links in a population of wild baboons studied by the Amboseli Baboon Research Project. The Amboseli Baboons Research Project has been collecting behavioral and demographic data on the Amboseli baboons for over 50 years, fecal hormone data for over 20 years, and blood samples collected via brief anaesthetizations for nearly 10 years. We complemented these remarkable datasets with cross-sectional data of female baboon body size.

In Chapter 1, we address two gaps in our understand of female dominance rank: (1) do higher-ranking females experience fewer stressors than lower-ranking females, and (2) how should we best quantify female dominance rank? Using fecal glucocorticoid concentrations as a proxy for the intensity and/or frequency of stressors that a baboon experiences, we find that, indeed, higher-ranking females do experience fewer stressors than lower-ranking females. Surprisingly, we also find that the best way to understand this effect is by categorizing females into two groups: alpha females, who are the highest-ranking female in the group, and everyone else.

In Chapter 2, we then focus on differences in the competitive landscapes assumed by two common measures of dominance rank, ordinal and proportional ranks. We complement theoretical work with re-analysis of 20 prior Amboseli baboon studies to show that for males, ordinal rank (i.e., number of individual ranking above the focal animal) was always a better predictor of traits than proportional rank, whereas for females, some traits were better predicted by ordinal rank, and some were better predicted by proportional rank (i.e., proportion of the group that a focal animal dominates). Our results suggest that males compete for density-dependent resources, whereas females compete for a mix of density-dependent and density-independent resources. In addition, our study demonstrates a new way to learn about the nature of within-group competition.

In Chapter 3, we present two new methods to use with body size data collected via parallel-laser photogrammetry. One of these methods was developed by colleagues here at Duke University, and the other method was developed by colleagues at George Washington University. These methods automate part of the hand-measurement process – measuring the distance between the lasers – and effectively saves time while increasing accuracy and precision of the final body size measurement. Our two methods have different strengths and weaknesses, and we anticipate that researchers will gravitate toward one or the other depending on their dataset, with the ultimate goal of increasing the use, ease, and accuracy of parallel-laser photogrammetry in studies of behavioral ecology.

In Chapters 4, we use the method developed in Chapter 3 to test whether early-life adversity stunts body size in female baboons. While this effect has been found in humans and some nonhuman animals, data on inter-individual differences in body size are extremely rare in wild primates. Using a dataset of over 2,000 images of 127 female baboons, we present the first cross-sectional growth curve of wild female baboons from juvenescence throughout adulthood. We then test whether females exposed to three main sources of early-life adversity - drought, maternal loss, or a cumulative measure of adversity – are smaller for their age in juvenescence or adulthood. We find that early-life drought predicts smaller limb length but not smaller torso length; our other measures of early-life adversity do not predict differences in body size. Our results suggest that baboons grow plastically in response to energetic early-life stress, but that this plasticity seems limited to limb growth, not torso growth.

Finally, in Chapter 5, we test a component of the biological embedding hypothesis, which predicts that early-life adversity is associated with elevated baseline inflammation as well as heightened acute inflammation in adulthood. To our knowledge, these predictions have only been tested in humans. Using serum samples collected from 89 baboons via brief anaesthetization, we measured several biomarkers of baseline and acute inflammation: c-reactive protein, soluble urokinase plasminogen activator receptor, interleukin 6, interleukin 1-beta, and tumor necrosis factor alpha. We test two measures of early-life adversity: maternal loss and a cumulative measure that incorporates 5 different potential sources of adversity. In contrast to the predictions of the biological embedding hypothesis, we find that baboons who experienced early-life adversity have a mix of comparable or lower levels of baseline and acute inflammation compared to baboons who experience no adversity. Prior tests of the biological embedding hypothesis were performed in humans who generally had access to more calories, less active lifestyles, and lower pathogen burden than wild baboons. Our results highlight the varied effects that early-life adversity can have on an organism’s development depending on the broader environment in which that organism lives.

Increasingly, drone-based photogrammetry has been used to measure size and body condition changes in marine megafauna. A broad range of platforms, sensors, and altimeters are being applied for these purposes, but there is no unified way to predict photogrammetric uncertainty across this methodological spectrum. As such, it is difficult to make robust comparisons across studies, disrupting collaborations amongst researchers using platforms with varying levels of measurement accuracy.

In this dissertation, I evaluate the major drivers of photogrammetric error and develop a framework to easily quantify and incorporate uncertainty associated with different UAS platforms. To do this, I take an experimental approach to train a Bayesian statistical model using a known-sized object floating at the water’s surface to quantify how measurement error scales with altitude for several different drones equipped with different cameras, focal length lenses, and altimeters. I then use the fitted model to predict the length distributions of unknown-sized humpback whales and assess how predicted uncertainty can affect quantities derived from photogrammetric measurements such as the age class of an animal (Chapter 1). I also use the fitted model to predict body condition of blue whales, humpback whales, and Antarctic minke whales, providing the first comparison of how uncertainty scales across commonly used 1-, 2-, and 3-dimensional (1D, 2D, and 3D, respectively) body condition measurements (Chapter 2). This statistical framework jointly estimates errors from altitude and length measurements and accounts for altitudes measured with both barometers and laser altimeters while incorporating errors specific to each. This Bayesian statistical model outputs a posterior predictive distribution of measurement uncertainty around length and body condition measurements and allows for the construction of highest posterior density intervals to define measurement uncertainty, which allows one to make probabilistic statements and stronger inferences pertaining to morphometric features critical for understanding life history patterns and potential impacts from anthropogenically altered habitats. From these studies, I find that altimeters can greatly influence measurement predictions, with measurements using a barometer producing larger and greater uncertainty compared to using a laser altimeter, which can influence age classifications. I also find that while the different body condition measurements are highly correlated with one another, uncertainty does not scale linearly across 1D, 2D, and 3D body condition measurements, with 2D and 3D uncertainty increasing by a factor of 1.44 and 2.14 compared to 1D measurements, respectively. I find that body area index (BAI) accounts for potential variation along the body for each species and was the most precise body condition measurement.

I then use the model to incorporate uncertainty associated with different drone platforms to measure how body condition (as BAI) changes over the course of the foraging season for humpback whales along the Western Antarctic Peninsula (Chapter 3). I find that BAI increases curvilinearly for each reproductive class, with rapid increases in body condition early in the season compared to later in the season. Lactating females had the lowest BAI, reflecting the high energetic costs of reproduction, whereas mature whales had the largest BAI, reflecting their high energy stores for financing the costs of reproduction on the breeding grounds. Calves also increased BAI opposed to strictly increasing length, while immature whales may increase their BAI and commence an early migration by mid-season. These results set a baseline for monitoring this healthy population in the future as they face potential impacts from climate change and anthropogenic stresses. This dissertation concludes with a best practices guide for minimizing, quantifying, and incorporating uncertainty associated with photogrammetry data. This work provides novel insights into how to obtain more accurate morphological measurements to help increase our understanding of how animals perform and function in their environment, as well as better track the health of populations over time and space.