Modelling of selected forest geometric parameters from airborne laser scanning data Abstract The main aim of this work is to approximate the shape of a tree crown with mathematically describable 3D shape based on airborne laser scanning (ALS) data. And consequently derive geometrical parameters describing the tree from this model. Included in the work is a custom designed algorithm based on angular segmentation. Measured results of this algorithm are then compared to an algorithm based on RANSAC and field measurement. The first part of this work describes airborne laser scanning, its use to derive characteristics of forest stands and individual trees and the theory of tree crown modelling. The next part contains a description of both algorithms and presentation of results and field measurements. The conclusion summarizes and evaluates the outputs of the custom angular segmentation algorithm and discusses its possible modifications. Keywords: airborne laser scanning, tree height, crown width, crown height, crown cover, crown volume, crown shape, RANSAC

Coregistration of airborne laser scanning data and aerial images Abstract This thesis is dealing with the co-registration of aerial laser scanning and aerial images. Theoretical part with research of current methods puts emphasis on methods suitable for remote sensing datasets. Part of the thesis is about pre-processing data for co-registration and DSM production. Selection of co- registration methods for remote sensing is based on previous researches. Selected co-registration methods are applied on datasets from EuroSDR research project and ČÚZK dataset. Application is realised by programming codes and functions that were created for this purpose in Matlab. Possibilities of usage, advantages and disadvantages of methods are being mentioned in the next parts of the thesis with emphasis on time of the computation and final accuracy. The function programmed in Matlab allows comparison of co-registration methods and allows the user to decide which of the co- registration methods to use on input datasets. Discussion section describes the possibilities of method extensions and problematic parts across the whole co-registration process. Keywords: co-registration, laser, scanning, images, photogrammetry, remote sensing, coordinate, image matching.

As regiões de cultivo de eucalipto apresentam um grande número de variações ambientais em nível microrregional, como pode ser facilmente constatado quando se observam as diferenças expressivas entre as produtividades florestais de cada área. No entanto, para se avaliar um ambiente, é necessário identificar segmentos específicos dentro de uma paisagem, tratando-se de uma propriedade florestal e, dentro desse estudo, buscar a compreensão das interações que aí se processam. Este trabalho contribui para entender o efeito da face de exposição solar no crescimento de plantios comerciais de Eucalyptus. A proposta metodológica apresentada leva em consideração o conhecimento do meio físico - tendo como base a fisiografia para distinção das faces de exposição solar- além disso, a interação dos fatores fisiográficos com os fatores biológicos são discutidos para subsidiar a tomada de decisão. A acentuada interação genótipo x ambiente presente em muitas culturas faz com que estudos de adaptabilidade a ambientes específicos sejam parte integrante dos programas de pesquisa florestal. A resposta diferenciada dos genótipos a ambientes favoráveis e desfavoráveis pode auxiliar o gestor florestal na alocação assertiva dos materiais genéticos para maximizar a produtividade. O escaneamento a laser aerotransportado, ferramenta de uso precoce no Brasil, é um sensor óptico ativo que mede a distância até os objetos-alvo e foi utilizado para a obtenção da altura das árvores. A área de estudo está localizada na região sudoeste do estado de São Paulo. Dentre as métricas disponíveis do processamento utilizou o percentil 90 de altura para estimar a altura de plantios comerciais homogêneos de Eucalyptus. Elaborou-se a análise de variância considerando um delineamento fatorial contemplando a face de exposição solar (norte e sul) e o material genético (clones comerciais), além da interação entre ambos para observar o efeito da face no crescimento em altura do povoamento. Para florestas abaixo de quatro anos de idade observou-se que os plantios de face sul cresceram menos que os da face norte. Para florestas acima de 5 anos essa diferença entre face continua, contudo, quando concluído o desdobramento da interação fatorial observou-se que os materiais genéticos plantados em face norte não diferem entre si, mas a diferença se mantém para florestas em face sul. / The eucalyptus cultivation regions boast a large number of environmental variables in micro-regional level, as can easily be seen when studying the expressive differences between the productivity of each forest area. To evaluate an environment, however, it is necessary to identify specific segments within a landscape with regards to a forest property and within that landscape study to seek the understanding of the interactions that takes place in the process. This work contributes to a better understanding of the effect of the aspect on growth in commercial plantations of Eucalyptus. The methodological approach presented in this work takes into account the knowledge of the physical environment, based on physiography, for distinction of aspect on environments, in addition to showing the interaction of physiographic factors with biological factors to support the decision-making process. The sharp interaction genotype x environment present in many cultures leads to studies of adaptability to specific environments to become integral parts of forestry research programs. The differentiated response of genotypes to favorable and unfavorable environments can assist forest managers in the allocation of assertive genetic materials to maximize productivity. The airborne laser scanning, which is an early tool used in Brazil, is an active optical sensor that measures the distance to the target objects, was used to obtain the height of the trees. The study area is located in the southwestern region of the State of São Paulo in Brazil. Among the available processing metrics, it was used 90 percentiles in height to estimate the height of homogeneous commercial plantations of Eucalyptus. A variance analysis was devised considering a factorial design contemplating the aspect (North and South) and genetic material (commercial clones), besides the interaction between both to observe the effect of the exposure on the height growth of the settlement. To forests below four years of age, it was observed that the plantations of the South exposure grow less than the north exposure. For forests above five years, this difference between aspect continues, however, when the deployment of factorial interactions is done, it\'s observed that the genetic materials planted in the north exposure do not differ with each other, but the difference stands within forests in the South exposure.

As regiões de cultivo de eucalipto apresentam um grande número de variações ambientais em nível microrregional, como pode ser facilmente constatado quando se observam as diferenças expressivas entre as produtividades florestais de cada área. No entanto, para se avaliar um ambiente, é necessário identificar segmentos específicos dentro de uma paisagem, tratando-se de uma propriedade florestal e, dentro desse estudo, buscar a compreensão das interações que aí se processam. Este trabalho contribui para entender o efeito da face de exposição solar no crescimento de plantios comerciais de Eucalyptus. A proposta metodológica apresentada leva em consideração o conhecimento do meio físico - tendo como base a fisiografia para distinção das faces de exposição solar- além disso, a interação dos fatores fisiográficos com os fatores biológicos são discutidos para subsidiar a tomada de decisão. A acentuada interação genótipo x ambiente presente em muitas culturas faz com que estudos de adaptabilidade a ambientes específicos sejam parte integrante dos programas de pesquisa florestal. A resposta diferenciada dos genótipos a ambientes favoráveis e desfavoráveis pode auxiliar o gestor florestal na alocação assertiva dos materiais genéticos para maximizar a produtividade. O escaneamento a laser aerotransportado, ferramenta de uso precoce no Brasil, é um sensor óptico ativo que mede a distância até os objetos-alvo e foi utilizado para a obtenção da altura das árvores. A área de estudo está localizada na região sudoeste do estado de São Paulo. Dentre as métricas disponíveis do processamento utilizou o percentil 90 de altura para estimar a altura de plantios comerciais homogêneos de Eucalyptus. Elaborou-se a análise de variância considerando um delineamento fatorial contemplando a face de exposição solar (norte e sul) e o material genético (clones comerciais), além da interação entre ambos para observar o efeito da face no crescimento em altura do povoamento. Para florestas abaixo de quatro anos de idade observou-se que os plantios de face sul cresceram menos que os da face norte. Para florestas acima de 5 anos essa diferença entre face continua, contudo, quando concluído o desdobramento da interação fatorial observou-se que os materiais genéticos plantados em face norte não diferem entre si, mas a diferença se mantém para florestas em face sul. / The eucalyptus cultivation regions boast a large number of environmental variables in micro-regional level, as can easily be seen when studying the expressive differences between the productivity of each forest area. To evaluate an environment, however, it is necessary to identify specific segments within a landscape with regards to a forest property and within that landscape study to seek the understanding of the interactions that takes place in the process. This work contributes to a better understanding of the effect of the aspect on growth in commercial plantations of Eucalyptus. The methodological approach presented in this work takes into account the knowledge of the physical environment, based on physiography, for distinction of aspect on environments, in addition to showing the interaction of physiographic factors with biological factors to support the decision-making process. The sharp interaction genotype x environment present in many cultures leads to studies of adaptability to specific environments to become integral parts of forestry research programs. The differentiated response of genotypes to favorable and unfavorable environments can assist forest managers in the allocation of assertive genetic materials to maximize productivity. The airborne laser scanning, which is an early tool used in Brazil, is an active optical sensor that measures the distance to the target objects, was used to obtain the height of the trees. The study area is located in the southwestern region of the State of São Paulo in Brazil. Among the available processing metrics, it was used 90 percentiles in height to estimate the height of homogeneous commercial plantations of Eucalyptus. A variance analysis was devised considering a factorial design contemplating the aspect (North and South) and genetic material (commercial clones), besides the interaction between both to observe the effect of the exposure on the height growth of the settlement. To forests below four years of age, it was observed that the plantations of the South exposure grow less than the north exposure. For forests above five years, this difference between aspect continues, however, when the deployment of factorial interactions is done, it\'s observed that the genetic materials planted in the north exposure do not differ with each other, but the difference stands within forests in the South exposure.

Beståndsindelning av skog är till stor den en manuell process som kräver mycket tid. De senaste 20 åren har tekniker som Airborne Laser Scanning (ALS) bidragit till en effektivisering av processen genom att generera laserdata som möjliggör skapandet av lättolkade bilder av skogsområden. Ur laser- och bilddata kan skogliga attribut så som trädhöjd, trädtäthet och markhöjd extraheras. Studiens syfte var att utvärdera vilka attribut som var mest relevanta för att särskilja skogsbestånd i ett system som delade in skog i bestånd automatiskt. Vid analys av attributens relevans användes klassificeringsmodeller. Fackmän intervjuades och litteratur studerades. Under studien modifierades systemets algoritmer med ambitionen att höja dess resultat till en tillfredsställande nivå. Studien visade att attribut som är kopplade till skogssköstel har störst relevans vid automatisk beståndsindelning. Trots modifieringar och använding av relevanta attribut lyckades studien inte påvisa att systemet kunde fungera som en egen lösning för beståndsindelning av skog. Däremot var den resulterande beståndsindelningen lämplig att använda som ett komplement vid manuell beståndsindelning.

Assessment of forest damage using satellite and lidar data Abstract The main objective of this thesis is to create a methodical procedure used for the evaluation of forest damage in the chosen area of the National Park Sumava, Czech Republic. In this work were combined the multispectral satellite data and data of airborne laser scanning. The forests in this area are heavily damaged mainly due to bark beetle outbreak. You can find here as healthy so damaged forests. Based on this methodology will be differentiated greater number of classes than I found in the literature. In this work was used pansharpened multispectral image SPOT, multispectral image Landsat and airborne laser scanning data with low density points. Another task was to get height information from ALS data in the form of grid. Forest stands were classified using object-oriented classification, which included at first segmentation and then creation of classification base. In classification entered spectral information and height information obtained from the ALS data. Forests were classified into 5 classes and accuracy of both classifications was evaluated using the error matrix and kappa coefficient. SPOT image classification reached kappa coefficient of 68,5 % and Landsat image classification reached kappa coefficient of 72,3 %. From the...

När kraven på mätosäkerhet är hög vid geodetiska mätningar behövs geodetiska referenssystem realiserade av geodetiska nät av hög kvalitet. Etableringen och transformation till överordnade referenssystem i höjd av dessa nät genomförs idag ofta med terrestra metoder som avvägning, vilket är ett noggrant men tidskrävande arbete. Det finns flera försök att använda sig av andra metoder såsom GNSS, men en möjlighet skulle även vara att använda punktmoln från flygburen laserskanning. Detta arbetes syfte är att undersöka om punktmoln kan användas till transformation av ett geodetiskt höjdnät i Sandvikens kommun. Nätet består av cirka 500 fixpunkter och har 2010 transformerats till RH2000 av Lantmäteriet. Det använda punktmolnet har producerats av Lantmäteriet och har en medelavvikelse om 0,05 m på plana hårdgjorda ytor. Detta är relativt högt då vanligen osäkerheter på millimeternivå önskas vid transformationer. Men eftersom medeltal kan reducera slumpmässiga avvikelser i enskilda mätningar kan en transformation bestående av ett medelhöjdskift möjligen ge ett tillfredsställande resultat. Medelhöjdskiftet är då ett medeltal av flera höjdskift beräknade på olika ställen i punktmolnet. Genom att avväga höjdskillnaden mellan fixpunkterna i nätet och punkter på markytan som med olika metoder höjdbestäms utifrån punktmolnet har höjder för fixpunkterna enligt punktmolnet erhållits. Dessa har jämförts med RH2000 höjder enligt Lantmäteriets transformation för att beräkna en avvikelse, samt med de äldre lokala höjderna för att beräkna ett höjdskift. Genom att beräkna medelvärde och dess osäkerhet för höjdskiftet över hela nätet har en uppfattning om metodens lämplighet erhållits. Höjdskiften låg överlag inom några millimeter från det som Lantmäteriet beräknat, med 3 mm osäkerhet för den överlag bästa metoden. Även om höjdskiftets och därmed transformationens osäkerhet delvis blev något hög jämfört med avvägning så kan punktmoln ändå vara ett lämpligt alternativ. Speciellt i mera avlägsna områden utan bra anknytningar till det överordnade nätet kan det vara intressant. Det finns dock många parametrar som ännu inte utforskats, bland annat vissa eventuella systematiska avvikelser. / When the requirements on accuracy and precision are high for geodetic measurements you need geodetic reference systems realized with geodetic control networks of high quality. Today, establishment and transformation to higher order reference systems for height usually uses terrestrial methods like levelling. While highly accurate these result in time consuming work. There have been a couple attempts at using other methods for this task, for example GNSS, but another possibility might be usage of point clouds from airborne laser scanning. As a starting point for further studies this study attempts to use point clouds to transform a geodetic height network in Sandviken municipality, Sweden. The network consists of around 500 benchmarks and has been transformed to the national reference system for height, RH2000, by the Swedish national geodetic survey (Lantmäteriet) in 2010. The point cloud used is also produced by Lantmäteriet and is said to have a mean error of 0,05 m. This is relatively high since the requirements usually are in the millimeter range when determining transformation parameters, but if the transformation only consist of a single height shift calculated as a mean from several height shifts derived from the point cloud any random errors in the point cloud should be reduced. By measuring the height difference between benchmarks and points on the ground, that through different methods are given heights according to the point cloud, heights of the benchmarks have been determined according according to the point cloud. These can be compared to heights in RH2000 according to the transformation performed by Lantmäteriet to see their deviation from the assumed true value. Further comparisons against the older local heights of the benchmarks give a height shift that can be used as a simple transformation. By calculating a mean and uncertainty an estimation of the suitability of the method can be achieved. The all height shifts deviated a few millimeters from the result Lantmäteriet got, with uncertainties around 3 mm for the overall best method. Even if the uncertainty of the shift and therefore the transformation ended up somewhat high compared to what Lantmäteriet achieved it is still believed that point clouds may be or become a viable alternative. Especially in more remote regions without good connections to the higher order network. There are many parameters that have not yet been explored though, as well as some potential systematic errors that should be further investigated.

This diploma thesis deals with the classification of the point clouds taken by different carrier and with various density. The terrain model and building models were created from provided data sets. Also the software equipment is described. Achieved outcomes of elaboration are presented in each corresponding chapter. In conclusion, the overall evaluation and assessment of the results of processing is done.

Flygburen laserskanning är en effektiv metod för insamling av höjddata över stora områden och används därför frekvent som underlag till digitala höjdmodeller, både på nationell och regional nivå (Wehr & Lohr 1999). Fördelen med insamlingsmetoden är att de utsända laserpulserna reflekteras på både markytan och objekten ovan mark, exempelvis vegetation, byggnader och liknande. På så vis genereras ett tredimensionellt punktmoln från vilket ytterligare produkter kan genereras. Den uppskattade eller uppmätta kvaliteten hos LiDAR-data gäller generellt för hela skanningsområdet. Men det kan vara av intresse att utföra en mer djupgående analys av kvaliteten för att se hur den skiljer sig mellan olika terrängtyper. På uppdrag av Arvika kommun ska en kvalitetskontroll av Lantmäteriets andra rikstäckande laserskanning ”Laserdata Skog” utföras. I dagsläget arbetar man med Laserdata NH, syftet med studien är således att ge Arvika kommun en mer nyanserad uppfattning av kvaliteten hos Laserdata Skog så att framtida arbeten kan ske på ett tillförlitligt sätt med en djupare förståelse kring datat. En jämförelse med Lantmäteriets första rikstäckande laserskanning ”Laserdata NH” kommer även utföras. Jämförelsen mellan de två laserskanningarna sker främst av ett teoretiskt intresse för att utreda hur stor skillnaden är mellan dem, framtida arbeten med laserdata kommer troligtvis att ske med den nya ”Laserdata Skog”.   För att utföra studien tillämpas den tekniska specifikationen SIS-TS 21144:2016 ”Byggmätning – Specifikationer vid framställning och kontroll av digitala markmodeller”. Inmätning av referensdata utfördes med både GNSS-utrustning och totalstation. De terrängtyper som har inkluderats i studien är: asfaltsyta, grusyta, lövskog, barrskog och gräsyta. För varje terrängtyp selekterades 2 provytor för att uppnå en god representation av de enskilda terrängtyperna. För att möjliggöra en koordinatjämförelse mellan laser- och referensdata så interpolerades punktmolnet till en TIN-yta. Resultatet visar att det uppstår differenser mellan Laserdata NH och Laserdata Skog för de statistiska mått som har beräknats. Laserdata NH erhåller förvånansvärt låga avvikelser. En övergripande trend är dock att Laserdata Skog har de lägre avvikelserna. Att fastställa orsakerna till dessa är dock svårt då det finns ett flertal faktorer som spelar in. Sammanfattningsvis erhåller Grusyta det lägsta RMSE-värdet (0,021 m) i Laserdata NH och i Laserdata Skog är det Asfaltsyta (0,017 m). Det högsta RMSE-värdet hittas i Barrskog för både Laserdata NH (0,198 m) och Skog (0,111 m). / Airborne laser scanning is an efficient method for collecting elevation data over a large area and is therefore frequently used as a basis for digital elevation models, both on a national and regional level (Wehr & Lohr 1999). The advantage of this data collection method is that the emitted laser pulses are reflected both on the ground surface as well as the objects above it, for example the vegetation, buildings or the like. In this way a three-dimensional point cloud can be created from which further products can be generated. The estimated or measured quality of LiDAR data generally applies for the entire scanning area. But it can be interesting to perform a more in-depth analysis of how the quality differs between different types of terrain.  At the request of Arvika municipality a quality survey of Lantmäteriet’s second nationwide laser scanning “Laserdata Skog” will be performed. Work is currently being performed using Laserdata NH, the purpose of this study is thus to give Arvika kommun a more nuanced perception of Laserdata Skog’s quality so that future work can be done in a more reliable way with a deeper knowledge about the data at hand. A comparison between Lantmäteriet’s first nationwide laser scanning “Laserdata NH” will also be performed. The comparison between these two is primarily out of a theoretical interest to examine how the quality differs between them. Future laserdata work will probably be executed using the newer product “Laserdata Skog”. The technical specification SIS-TS 21144:2016 ”Construction measurements – Specifications of production and control of digital terrain models” was applied in the study. Both GNSS-equipment and total station where used in order to collect reference data. The included terrain types are: asphalt, gravel, deciduous forest, coniferous forest and grass. Two areas of interest have been selected for each type of terrain in order to achieve a good representation of each terrain type. In order to perform a coordinate comparison between the laser- and reference data the point cloud from the laserdata was interpolated to a TIN-surface. The results show that there are quality differences between Laserdata NH and Laserdata Skog. Laserdata NH obtains remarkably low deviations. The overall trend is however that Laserdata Skog acquires the lower deviations of the two. Determining the causes of this is difficult, as there are several factors that come in to play. In summary the Gravel terrain type obtains the lowest RMSE-value (0,021 m) for Laserdata NH. The terrain type with the lowest RMSE-value for Laserdata Skog is Asphalt (0,017 m). The highest RMSE-values are found in Coniferous forest for both Laserdata NH (0,198 m) and Laserdata Skog (0,111 m).

Höjddata av god kvalitet är av stor betydelse i många sammanhang, inte minst i samhällets anpassning till ett förändrat klimat. Laserdata Skog är höjddata insamlade från flygburen laserskanning och kommer i färdigt skick att täcka större delen av Sveriges yta. För att sådana data ska kunna användas på bästa sätt är det viktigt att ha kännedom om deras kvalitet. För data från flygburen laserskanning finns det många parametrar som orsakar variation i kvaliteten, där terrängtypens inverkan hör till de mest betydande.  Denna studie utförs på uppdrag av Lantmäteriet och syftar till att undersöka kvaliteten i Laserdata Skog. Fokus ligger på osäkerhet i höjd och punkttäthet, samt hur dessa faktorer varierar mellan olika terrängtyper. Höjdosäkerheten har undersökts genom jämförelser mellan laserdata och terrestra kontrollmätningar, medan punkttätheten har bestämts med beräkningar och observationer i laserdata. Fyra terrängtyper ingår i studien: Hårdgjord yta, Gräsyta, Barrskog och Lövskog. Varje terrängtyp representeras av tre olika provytor, fördelade på olika skanningsområden. Osäkerheten i höjd påverkades av både trädskikt och markvegetation, medan det enbart var variation i trädskiktet som orsakade synbar påverkan på punkttätheten. Osäkerheten i höjd för enskilda provytor varierade mellan 0,011 och 0,183 m (RMS). Punkttätheten varierade mellan 0,66 och 2,09 punkter/m2. För osäkerheten i höjd påträffades ett betydande bidrag från inpassningen av punktmolnet, vilket försvårade analysen av terrängtypens inverkan. / High quality elevation data is of great importance in many contexts, for example in society’s adaptation to climate change. Laser data forest (Laserdata Skog) is elevation data collected from airborne laserscanning, and will cover most of Sweden’s surface when completed. In order for this data to be used in the best possible way, knowledge of its quality is important. Many parameters causes variation in quality for airborne laserdata, and the impact of vegetation is one of the most significant. This study is conducted by request from Lantmäteriet (the Swedish mapping, cadastral and land registration authority) and aims to investigate the quality of Laser data Forest. Uncertainty in height and point density is the main focus, as well as how these factors vary in different types of terrain. Uncertainty in height has been investigated by comparisons between laser data and terrestrial control measurements, while point density has been determined by calculations and observations in laser data. Four types of terrain is included in the study: Impervious surface, Grass, Coniferous forest and Deciduous forest. Each type of terrain is represented by three test surfaces, one in each of three different scanning areas. Uncertainty in height was affected by both trees and ground cover, while the vegetational impact on point density was caused by trees alone. Uncertainty in height for individual test sites varied between 0,011 and 0,183 m (RMS). Point density varied between 0,66 and 2,09 points/m2. For the uncertainty in height, a considerable contribution was found to originate from the alignment of the point clouds, which made the analysis of the impact of the terrain more difficult.