The quality of indirect measurements of canopy structure, from in situ and satellite remote sensing, is based on knowledge of vegetation canopy architecture. Technological advances in ground-based, airborne or satellite remote sensing can now significantly improve the effectiveness of measurement programs on forest resources.The structure of vegetation canopy describes the position, orientation, size and shape of elements of the canopy.The complexity of the canopy in forest environments greatly limits our ability to characterize forest structural attributes. Architectural models have been developed to help the interpretation of canopy structural measurements by remote sensing. Recently, the terrestrial LiDAR systems, or TLiDAR ( Terrestrial Light Detection and Ranging ), are used to gather information on the structure of individual trees or forest stands.The TLiDAR allows the extraction of 3D structural information under the canopy at the centimetre scale.The methodology proposed in my Ph.D. thesis is a strategy to overcome the weakness in the structural sampling of vegetation cover.The main objective of the Ph.D. is to develop an architectural model of vegetation canopy, called L-Architect (LiDAR data to vegetation Architecture ), and to focus on the ability to document forest sites and to get information on canopy structure from remote sensing tools. Specifically, L-Architect reconstructs the architecture of individual conifer trees from TLiDAR data. Quantitative evaluation of L-Architect consisted to investigate (i) the structural consistency of the reconstructed trees and (ii) the radiative coherence by the inclusion of reconstructed trees in a 3D radiative transfer model. Then, a methodology was developed to quasi-automatically reconstruct the structure of individual trees from an optimization algorithm using TLiDAR data and allometric relationships. L-Architect thus provides an explicit link between the range measurements of TLiDAR and structural attributes of individual trees. L-Architect has finally been applied to model the architecture of forest canopy for better characterization of vertical and horizontal structure with airborne LiDAR data. This project provides a mean to answer requests of detailed canopy architectural data, difficult to obtain, to reproduce a variety of forest covers. Because of the importance of architectural models, L-Architect provides a significant contribution for improving the capacity of parameters' inversion in vegetation cover for optical and lidar remote sensing.

This research aims to further the understanding of ecological restoration success in the Tampa Bay, Florida, region. Although over four hundred restoration projects have been completed in the bay area, knowledge of their success has been hindered by the lack of assessment and transfer of information concerning project outcomes. Without comprehensive project assessment, local science will be limited in its ability to inform practice because we lack the advantage of past knowledge. Using a multi-scaled approach, a diverse set of restoration projects are evaluated, providing information on how the projects are contributing to defined targets established by the Tampa Bay Estuary Program’s guiding documents. Through execution of habitat field assessments and completion of geographic information system, remote sensing, and aerial and terrestrial laser scanning analyses, the feasibility and effectiveness of these projects is investigated. Additionally, the research provides innovative techniques for monitoring projects with relative ease, allowing project evaluation to be conducted on a more regular basis across a range of temporal and spatial scales. A cost matrix, created from this toolbox, is provided to offer land managers with a means of evaluating, regulating, and conserving restored critical coastal habitats in Tampa Bay, thus saving public dollars that may otherwise be wasted on failed projects.

This dissertation combines three separate studies that measure coastal change using airborne laser data. The initial study develops a method for measuring subaerial and subaqueous volume change incrementally alongshore, and compares those measurements to shoreline change in order to quantify their relationship in Palm Beach County, Florida. A poor correlation (R2 = 0.39) was found between shoreline and volume change before the hurricane season in the northern section of Palm Beach County because of beach nourishment and inlet dynamics. However, a relatively high R2 value of 0.78 in the southern section of Palm Beach County was found due to little disturbance from tidal inlets and coastal engineering projects. The shoreline and volume change caused by the 2004 hurricane season was poorly correlated with R2 values of 0.02 and 0.42 for the north and south sections, respectively. The second study uses airborne laser data to investigate if there is a significant relationship between shoreline migration before and after Hurricane Ivan near Panama City, Florida. In addition, the relationship between shoreline change and subaerial volume was quantified and a new method for quantifying subaqueous sediment change was developed. No significant spatial relationship was found between shoreline migration before and after the hurricane. Utilization of a single coefficient to represent all relationships between shoreline and subaerial volume change was found to be problematic due to the spatial variability in the linear relationship. Differences in bathymetric data show only a small portion of sediment was transported beyond the active zone and most sediment remained within the active zone despite the occurrence of a hurricane. The third study uses airborne laser bathymetry to measure the offshore limit of change, and compares that location with calculated depth of closures and subaqueous geomorphology. There appears to be strong geologic control of the depth of closure in Broward and Miami-Dade Counties. North of Hillsboro Inlet, hydrodynamics control the geomorphology which in turn indicates the location of the depth of closure.

The Province of British Columbia is developing a suite of attributes to assess and monitor forest sustainability. Each attribute is in turn evaluated using a variety of indicators. Recently, digital remote sensing technologies have emerged as both alternative and supplement to traditional monitoring techniques, with light detection and ranging (lidar) in particular showing great promise for estimating a variety of indicators. The goal of this thesis was to review and assess the ability of lidar to estimate selected indicators of forest sustainability. Specifically, digital elevation model (DEM) interpolation (from which indicators are extracted both directly and indirectly) and wildlife tree class distributions were examined. Digital elevation models are a key derivative of lidar data, and their generation is a critical step in the data processing stream. A validation exercise was undertaken to determine which combination of interpolation routine and spatial resolution was the most accurate. Ground returns were randomly subsetted into prediction and validation datasets. Linear, quintic, natural neighbour, spline with tension, regularized spline, inverse distance weighting, and ANUDEM interpolation routines were used to generate surfaces at spatial resolutions of 0.5, 1.0, and 1.5 m. The 0.5 m natural neighbour surface was found to be the most accurate (RMSE=0.17 m). Classification and regression tree analysis indicated that slope and ground return density were the best predictors of interpolation error. The amount and variability of living and dead wood in a forest stand is an important indicator of forest biodiversity. In the second study, the capacity of lidar to estimate the distribution of living and dead trees within forests is investigated. Twenty-two field plots were established in which each stem (DBH>10cm) was assigned to a wildlife tree (WT) class. For each plot, a suite of lidar-derived predictor variables were extracted. Ordinal logistic regression was then employed to predict the cumulative proportions of stems within the WT classes. Results indicated that the coefficient of variation of the lidar height data was the best predictor variable (r = 0.85, p <0.000, RMSE = 4.9%). The derived relationships allowed for the prediction of the proportion of stems within WT classes across the landscape. / Forestry, Faculty of / Graduate

Sub-surface reservoirs are normally characterised by limited information from widely spaced wells (1D sections) and relative low-resolution seismic data (2D or 3D sections) making any derived geological model highly interpretive. The ability of outcrop exposures to study the sedimentary architecture (e.g. stacking patterns, lateral continuation, facies proportions and distribution) at a wide range of scales (km to mm) in continuous 3D accessible sections make them ideal analogues to help bridge the gap in resolution between seismic and well data in reservoir studies.In this study the sedimentary architecture of the outcrop exposures the Late Triassic Wolfville Formation are evaluated using traditional sedimentological (e.g. correlation panels, log sections) and digital (e.g. DGPS, LiDAR, DEM) field techniques to help better understand these often complex fluvial depositional systems for analogue reservoir studies. The sediments of the Wolfville Formation are superbly exposed in both cliff sections and on extensive wave-cut platforms along the shore of the Minas Basin (Nova Scotia, Canada). The succession lies unconformably on Pre-Triassic rocks and forms the earliest syn-rift unit in the Fundy basin. The unit comprises coarse- and fine-grained fluvial sandstones, aeolian dune deposits, and alluvial fan sediments. The laterally extensive and three-dimensional nature of the outcrop exposures offer a valuable insight in the sedimentary architecture of the fluvial system. A detailed sedimentological analysis of the succession allowed the determination of the large-scale sedimentary architecture of this gravelly to sandy bedload fluvial system. In order to characterise the architectural evolution in more detail digital outcrop studies were employed in three carefully selected study areas. The study areas are located along the southern Minas Basin shore within the gravelly and younger sandy-dominated part of the succession.The digital outcrop studies carried out in the study areas collected spatial data using differential GPS and LiDAR (Light Detection and Ranging) equipment. The integration of traditional collected geological and spatial data forms digital outcrop models (DOMs), which allow accurate mapping and evaluation of the geological properties geometry and distribution using novel spatial analysis techniques (e.g. classified point-cloud and perpendicular projection plane method). The results offer a better understanding on the heterogeneity of the depositional system, accurate geostatistical information on the characteristics of sedimentary bodies and genetic units (geometry, distribution and proportion) and provides a high-resolution stratigraphic-structural framework for geocellular outcrop models. In the geocellular outcrop models of the study areas the three-dimensional facies distribution are simulated using various modelling approaches, such as sequential indicator simulations, object modelling and multiple point statistics. The available control of the outcrop models allowed the various facies modelling approaches to be examined. The results are evaluated and discussed using qualitative comparison studies. In addition, these outcrop models provided detailed information on the three-dimensional fluvial architecture.

Achieving robot autonomy is an extremely challenging task and it starts with developing algorithms that help the robot understand how humans perceive the environment around them. Once the robot understands how to make sense of its environment, it is easy to make efficient decisions about safe movement. It is hard for robots to perform tasks that come naturally to humans like understanding signboards, classifying traffic lights, planning path around dynamic obstacles, etc. In this work, we take up one such challenge of motion segmentation using Light Detection and Ranging (LiDAR) point clouds. Motion segmentation is the task of classifying a point as either moving or static. As the ego-vehicle moves along the road, it needs to detect moving cars with very high certainty as they are the areas of interest which provide cues to the ego-vehicle to plan it's motion. Motion segmentation algorithms segregate moving cars from static cars to give more importance to dynamic obstacles. In contrast to the usual LiDAR scan representations like range images and regular grid, this work uses a modern representation of LiDAR scans using permutohedral lattices. This representation gives ease of representing unstructured LiDAR points in an efficient lattice structure. We propose a machine learning approach to perform motion segmentation. The network architecture takes in two sequential point clouds and performs convolutions on them to estimate if 3D points from the first point cloud are moving or static. Using two temporal point clouds help the network in learning what features constitute motion. We have trained and tested our learning algorithm on the FlyingThings3D dataset and a modified KITTI dataset with simulated motion.

The presented diploma thesis deals with the division of points into homogeneous groups. The work provides a broad overview of the current state in this topic and a brief explanation of the main segmentation methods principles. From the analysis of the articles are selected and programmed five algorithms. The work defines the principles of selected algorithms and explains their mathematical models. For each algorithm is also given a code design description. The diploma thesis also contains a cross comparison of segmentation capabilities of individual algorithms on created as well as on measured data. The results of the curves extraction are compared with each other graphically and numerically. At the end of the work is a comparison graph of time dependence on the number of points and the table that includes a mutual comparison of algorithms in specific areas.

Znalost o prostorovém rozložení, koncentraci a zdrojích CO2 v atmosféře je klíčová k pochopení přírodního cyklu oxidu uhličitého, k předpovědi vývoje a vlivu CO2 na klimatické změny. Tato práce se zabývá problematikou optického dálkového snímání za použití LIDAR (Light Detection and Ranging) systému. Obsahuje potřebné teoretické znalosti o LIDAR systému, použití a principy. Z mnoha aplikací využívající LIDAR je v této práci nastíněno provedení a měření pomocí DIAL (Differential Absorption LIDAR) systému určeného k určení koncentrace CO2 v atmosféře, tak i využití dalších aktivních či pasivních způsobů snímání CO2.

The objective of this study is to explore the methods used for automatic roof reconstruction so far and on the basis of this analysis purpose a methodology, which uses the geometric structure of Straight Skeleton for an initial approximation of the roof shape. In the first part of this thesis issues of automatic detection and extraction of building roof planes from laser scanning data are explored. Also, the Straight Skeleton is described in detail here and its potential for the construction of hip and saddle roofs is explained. An iterative approach which consists of deleting or moving appropriate points between roof planes using the principles of orthogonal regression is specified as an optimization method. In the second part of this thesis the proposed algorithm was implemented using CGAL (an open source library) and then it was tested on different data sets. In the very end, it is concluded that the use of the proposed algorithm on more complex types of roofs is inappropriate. The excellent results of the optimization of hip and saddle-shaped roof types are highlighted simultaneously.

Title: Automatic On-Line Calibration and Calibration Monitoring of Cameras and Lidars Author: Jaroslav Moravec Department: Department of Software and Computer Science Education Supervisor: doc. RNDr. Elena Šikudová, Ph.D., Department of Software and Computer Science Education Abstract: Cameras and LiDARs are important devices in the automotive indus- try as their combination provides useful information (3D coordinates of a point, its colour and intensity) for perception, localization, mapping and prediction. Successful data fusion and interpretation requires accurate calibration of intrin- sic parameters of the sensors and their 6D relative pose. In this thesis, we present a target-less calibration method on three different calibration tasks. The solu- tion is based on a robust likelihood function constructed over the reprojection error of LiDAR edges relative to image edges. When the calibration slowly wears off, our online recalibration procedure can jointly follow the extrinsic calibration drift with an average error of 0.13◦ in rotation and 4 cm in translation. Based on this recalibration tool, we are also able to monitor the calibration and detect an abrupt decalibration in a couple of seconds. And we also present a fully automatic calibration routine that estimates both the extrinsic and intrinsic...