This report summarizes the task of conceptually designing an UAV suited for agricultural observation of Swedish farmland. The design of the UAV was divided into two parts. This report focuses on the flight mechanics, performance analysis, and cost analysis of the UAV, whereas the other part centers around the aerodynamic performance. Therefore, some elements, such as the wing selection, will not be subject to discussion in this report. A set of different requirements were posed, such as having a flight time longer than two hours, being able to between 5-10 m/s, able to perform vertical take-off and landing, fly at a maximum of 100 meters, and weighing less than 5 kg. By using different sources of literature, reasonable assumptions, and Matlab analytics, a UAV was designed that met all constraints demanded. The cost analysis yielded a result that was reasonable, which overall makes this conceptual UAV a realistic product that could be manufactured using the project design.

Using autonomous underwater vehicles is a popular method of collecting samplesand conducting surveys, but the transportation of this information is not always easy.The underwater vehicle might be unable to transmit the information wirelessly, andsamples may be required to be transported a long distance. A possible solution tothese problems is a hybrid unmanned aerial vehicle, accompanying said underwatervehicle. After a submerged deployment, this vehicle could transport the informationover long distances, or conduct other operations at different locations in air or water.While quadcopters are an increasingly popular type of vehicle, conventional fixed­wingplanes are still superior when it comes to range. This thesis designs, builds and testssuch a vehicle, with the goal of a submerged deployment, performing vertical takeoff,and then transitioning to fixed­wing flight. To minimize the drone’s impact on thevehicle which it accompanies, it is nearly buoyancy neutral by flooding the hull withwater, which enters and exits the vehicle rapidly during dive and egress. To managethe pressure at the underwater vehicle’s operating depth, it utilizes a bladder ratherthan having a heavy rigid compartment. It floats as a tailsitter at the surface, usingtwo motors in a tractor configuration to pull itself out of the water. The vehicle builtproved capable of being submerged and taking off vertically, however there were nofixed­wing flights attempted. / Autonoma undervattensfordon är en populär metod för att samla in prover ochgöra undersökningar, men det är inte alltid uppenbart hur denna informationska transporteras. Undervattensfarkosten kanske inte har kapaciteten att sändainformationen trådlöst, och prover kan krävas att transporteras en lång sträcka. Enmöjlig lösning på dessa problem är en hybrid obemannad flygande farkost som följermed undervattensfordonet. Efter en undervattensstart kan detta fordon transporterainformationen över långa avstånd, alternativt genomföra andra operationer på olikaplatser. Quadcoptrar är en alltmer populär typ av fordon, men konventionella flygplanär fortfarande överlägsna när det gäller räckvidd. Denna avhandling konstruerar,bygger och testar ett sådant fordon, med målet att efter en undervattenstart, genomföravertikal start från vattenytan och sedan övergå till horisontell flygning. För attminimera drönarens inverkan på fordonet som den transporteras med har den nästanneutral flytkraft, detta sker genom att kroppen fylls med vatten, som snabbt kommerin och ut ur fordonet under dykning och vid flygning. För att hantera trycket påundervattensfordonets operationsdjup använder den en blåsa i stället för att ha enstyv behållare. Den flyter med svansen ner på ytan, och använder två motorer ien traktorkonfiguration för att dra sig ur vattnet. Fordonet som byggts visade sigkunna sänkas ned under vattnet och lyfta vertikalt, men det gjordes inga horisontellaflygningar.

Unmanned Aerial Systems (UAS) have become increasingly popular recently for surveying and mapping because of their efficiency in acquiring remotely sensed data in a short amount of time and the low cost associated with them. They are used to generate digital elevation models (DEM) derived from aerial photography for various purposes such as the documentation of cultural heritage sites, archaeological surveying or earthwork volume calculations. This thesis investigates the possible effects different file formats may have on the quality of elevation models. In this thesis, an UAS survey was simulated using a digital camera to produce six DEMs based on JPEG, TIFF and RAW format in Agisoft Photoscan by taking two sets of images of a city model, in different light conditions. Furthermore, a reference DEM was produced in Geomagic Studio using data from a Leica Nova MS50 Multistation. The DEMs were then compared in Geomagic Control. The results from the 3D comparison in Geomagic Control show that the standard deviation of all elevation models is 4 mm with the exception of the elevation model derived from raw-edited images taken with lighting, which has a standard deviation of nearly 6 mm. Also, all of the models have an average deviation of 0.4 mm or less. The significant deviations in all DEMs occur in areas where the multistation lacked vision of certain objects of the city model such as walls, or on the edges of the analysed area. Additionally, the georeferencing results from Photoscan show that the DEMs based on normal light condition images have slightly lower georeferencing errors than the DEMs with lighting. It has been concluded that it is difficult to say whether file formats have any noticeably effect on the uncertainty of digital elevation models.

The task of railroad inspection is a tedious one. It requires a lot of skilled experts and long hours of frequent on-field inspection. Automated ground equipment systems that have been developed to address this problem have the drawback of blocking the rail service during inspection process. As an alternative, using aerial imagery from a UAV, Computer Vision and Machine Learning based techniques were developed in this thesis to analyze two kinds of defects on the rail tracks. The defects targeted were missing spikes on tie plates and cracks on ties. In order to perform this inspection, the rail region was identified in the image and then the tie plate and tie regions on the track were detected. These steps were performed using morphological operations, filtering and intensity analysis. Once the tie plate was localized, the regions of interest on the plate were used to train a machine learning model to recognize missing spikes. Classification using SVM resulted in an accuracy of around 96% and varied greatly based on the tie plate illumination and ROI alignment for Lampasas and Chickasha subdivision datasets. Also, many other different classifiers were used for training and testing and an ensemble method with majority vote scheme was also explored for classification. The second category of learning model used was a multi-layered neural network. The major drawback of this method was, it required a lot of images for training. However, it performed better than feature based classifiers with availability of larger training dataset. As a second kind of defect, tie conditions were analyzed. From the localized tie region, the tie cracks were detected using thresholding and morphological operations. A machine learning classifier was developed to predict the condition of a tie based on training examples of images with extracted features. The multi-class classification accuracy obtained was around 83% and there were no misclassifications seen between two extreme classes of tie condition on the test data. / Master of Science

Yes / The offshore wind energy is increasingly becoming an attractive source of energy due to having lower environmental impact. Effective operation and maintenance that ensures the maximum availability of the energy generation process using offshore facilities and minimal production cost are two key factors to improve the competitiveness of this energy source over other traditional sources of energy. Condition monitoring systems are widely used for health management of offshore wind farms to have improved operation and maintenance. Reliability of the wind farms are increasingly being evaluated to aid in the maintenance process and thereby to improve the availability of the farms. However, much of the reliability analysis is performed offline based on statistical data. In this article, we propose a drone-assisted monitoring based method for online reliability evaluation of wind turbines. A blade system of a wind turbine is used as an illustrative example to demonstrate the proposed approach. / SURE Grant scheme. SESAME H2020 Project under Grant 101017258.

No description available.

A time-effective coverage path can be decisive in catastrophic and war scenarios for saving countless lives where UAVs are used to scan an area looking for an objective. Given an area shaped as a polygon, a quadratic decomposition method is used to discretize the area into nodes. A model of the optimization problem constraint is created and solved using mixed-integer linear programming, taking into consideration simple dynamics and coverage path planning definitions. Simulations in different scenarios are presented, showing that the presence of no-fly zones can negatively affect the coverage time. The relationship between coverage time and the number of UAVs employed is nonlinear and converges to a constant value. The result has a direct impact on the evaluation of benefits and the cost of adding UAVs to a search mission.

This research involved testing several aspects of industrial hemp (Cannabis sativa L.) production, including the impact of tillage on seed and fiber production, optimal harvest time for seed yield and quality, the response of seed yield to nitrogen fertility rates, and the potential of hemp as a forage crop. A three-year study was conducted in Blacksburg and Orange of Virginia State to assess the effects of tillage management and production systems (e.g., seed, dual, and fiber) on hemp establishment and productivity. Two cultivars, Joey (a dual-purpose variety) and EcoFibre (bred specifically for fiber), were planted into seedbeds prepared with conventional tillage and no-till management. The cultivar Joey, lower plant populations under seed production systems resulted in taller plants (P = 0.0002) compared to the dual-purpose production systems in 2020. Greater plant heights (P < 0.0001) with fiber production systems in 2021 and 2022 were due to differences between cultivars and their time of flowering. Conventional tillage resulted in greater (P ≤ 0.0161) plant populations than no-tillage for all production systems in each year, and this response was more pronounced with fiber management in 2020 (tillage × production systems interaction; P = 0.0007). Greater (P < 0.001) yields with fiber systems observed in 2021 and 2022 were largely driven by the more productive EcoFibre cultivar. Despite treatment differences in population density, biomass and seed yields varied less by tillage management and production systems. Lower plant population density was associated with greater biomass and seed yields per plant. However, for desired fiber quality and mechanical harvest feasibility, a higher plant population density is recommended. A second study aimed to determine the optimum harvest time for seed yield of two hemp cultivars. 'Joey', and 'Grandi,', were established in Blacksburg and Orange, Virginia in mid-May/early June of 2021 and 2022. The experiment was conducted as a randomized complete block design with a repeated measurement arrangement and four replicates. Plants were harvested four times at one-week intervals starting in mid-summer. Harvest date significantly affected seed yield, with the response differing by cultivar (cultivar × date interaction; P = 0.001) in 2022 at the Orange site. In Blacksburg, seed yields were similar for the two cultivars and greatest at the second harvest each season (July 22, 2021, and July 25, 2022), although they were substantially lower in 2022 due to drought (1750 vs. 480 kg ha-1; P < 0.0001). In Orange, in 2021, as planting occurred late, harvests were also deferred until August 17, and seed yields were greatest at this first harvest (1180 kg ha-1; P<0.0001). In 2022, yields at the Orange location were highest for Grandi at the first harvest (July 21; 1510 kg ha-1) and for Joey at the second harvest (July 28; 1280 kg ha-1) (Harvest Time by Cultivar interaction, P = 0.0010). Over the subsequent weeks of harvest, yields drastically declined (16 to 41% in 2021 and 27 to 47% in 2022 in Blacksburg; 52% to 91% in 2021 and 28% to 65% in 2022 in Orange, compared to the highest yield). Harvest timing is critical to achieving optimum seed yield, and it varies with cultivar, eco-physiographic location, and weather (e.g., rainfall). Fatty acids (FA) varied by cultivar, location, and harvest timing, but patterns of response were not consistent across FA. Gamma-linolenic (P ≤ 0.002) and oleic acids (P ≤ 0.023) were generally greater in Joey, with greater arachidic acid (P ≤ 0.013) concentrations in Grandi. Stearidonic acid concentrations declined with later harvest date in Orange location (P ≤ 0.0034). A third study aimed to measure hemp's response to different N rates and to determine the ability to predict plant N content and seed yield based on UAV-based multispectral imagery. Two hemp cultivars, 'Joey' and 'Grandi', were planted and five N rates (0, 60, 120, 180, 240 kg N ha-1) were tested in Blacksburg, Virginia in 2020, 2021, 2022. Aerial image acquisition occurred at three different growth stages in 2021 using dji M 300 drones mounted with multispectral sensors. Red/Blue index (R2=0.89), near-infrared (NIR) band (R2=0.84) and Enhanced vegetation index (EVI) (R2=0.81) were better predictors of N content in leaf samples than other vegetation indices that were evaluated. Green normalized difference vegetation index (GNDVI) was the better predictor of hemp seed yield (R2=0.58) than other evaluated vegetation indices. The seed yield of hemp was influenced (P ≤ 0.0177) by the N input in all three experimental years. In 2020, seed yield did not increase steadily with the increase of N rate; the highest seed yield, 1640 kg ha-1, was observed at 120 kg N ha-1. In 2021, maximum seed yield of 2500 kg ha-1 occurred at the maximum N rate (240 kg N ha-1). In 2022, a weak response to N rate was observed; maximum seed yield was 380 kg ha-1, again at 240 kg N ha-1. The overall growth of the hemp plants was affected by limited rainfall and weed pressures in 2022, leading to a significant reduction in seed yield. Response to N rate will vary depending on other factors such as available soil moisture during the growing season, weed pressure, and growing period. A fourth study examined the yield and nutritive value of three hemp cultivars, 'Grandi', 'Joey', and 'EcoFibre' as potential forage crops when harvested at weekly intervals in Blacksburg, VA. The greatest biomass and TDN yields across cultivars were 3.17 Mg ha-1 and 2.08 Mg ha-1 respectively, at two months after establishment in 2021. In the dry 2022 season, biomass and TDN yield were 1.9 Mg ha-1 and 1.03 Mg ha 1, respectively, two months after establishment. Hemp nutritive value measures varied by cultivar and harvest time (P < 0.05). Depending on the cultivar and harvest time, hemp plant biomass contained 13 to 32% CP, 22 to 45% NDF, 20 to 38% ADF, 4 to 9% lignin, and 52 to 80% TDN (cultivar × time interaction; P < 0.05). Hemp CP and TDN decreased gradually with maturation while ADF, NDF, and lignin increased (P<0.0001); however, this decline with maturity did not appear as severe as occurs with many other forages. These preliminary results suggest that hemp has the potential to be used as a forage crop. More research is needed to address hemp management and utilization, including field establishment and production, harvest timing for optimum tonnage and forage quality, and animal intake and performance studies. These findings provide new insights into industrial hemp production in the mid-Atlantic region of the United States. Optimal tillage practices, precise harvest timing, appropriate N fertility rates, and proper management techniques all are crucial for maximizing hemp seed and fiber production and quality. Furthermore, hemp shows promise as a forage crop with its adaptability and favorable nutritional properties. Further research is warranted to refine cultivation techniques, improve crop quality, and explore the full potential of hemp in various industries. / Doctor of Philosophy / Industrial hemp (Cannabis sativa L.) is a versatile crop with numerous applications in various industries, but much work must be done to understand crop responses to management practices and improve its potential as a crop for greater sustainability. In this study, we explored different aspects of hemp agronomic management. Hemp traditionally has been planted into tilled fields, which increases the chance for soil erosion. We examined whether hemp could be established without tillage and found that although tilled fields generally had great populations of taller plants; total biomass and seed yields were not as influenced by tillage. Our research suggests that with some tweaking, hemp can be successfully established without soil tillage. Next, we investigated the optimal time to harvest hemp for maximum seed yield. Harvesting at the right moment is crucial, as seeds ripen unevenly, resulting in varying quality and yield. By carefully timing the harvest, we can maximize seed yield and ensure high-quality seeds. Our cultivars were best harvested in a late July to early August time frame. Under favorable weather conditions, we observed seed yields ranging from 1,180 to 2,510 kilograms per hectare, depending on the hemp cultivar and location. Additionally, we studied the response of hemp seed yield to nitrogen fertilization rates. Nitrogen is an essential nutrient for plant growth, and we found that nitrogen significantly influenced seed yield, although the pattern of response varied by growing conditions. Over three years, seed yields ranged from 380 to 2,510 kilograms per hectare. Yields generally increased with nitrogen inputs but were highly affected by available moisture. Fertility studies help farmers determine the ideal nitrogen levels for their hemp crops, promoting healthy growth, maximizing yield, and minimizing environmental contamination. Within this study, we also evaluated aerial imagery technologies to monitor plant nitrogen status and we observed that remote sensing technologies are promising for building predictive nutrient management tools. Lastly, we explored the potential of hemp as a forage crop. Hemp plants have unique nutritional properties (e.g., protein, fatty acids) and can be used as feed for livestock. We investigated the best time to harvest hemp for maximum biomass and nutrient content, important factors for animal nutrition. Hemp plants grow rapidly and within two months after establishment they yielded up to 3.17 metric tons of biomass per hectare, with relatively high nutritional value. Overall, these studies provide valuable insights into hemp production, including the importance of tillage practices, optimal harvest timing, and appropriate nutrient management. By applying these findings, farmers can enhance their hemp cultivation techniques, resulting in higher yields, improved crop quality, and better environmental outcomes.

A new flying wing design has been developed at the Unmanned Systems Lab (USL) at Virginia Tech to serve delivery and remote sensing applications in the developing world. The fully autonomous unmanned aerial vehicle (UAV), named EcoSoar, was designed with the goal of creating a business opportunity for local entrepreneurs in low-resource communities. The system was developed in such a way that local fabrication, operation, and maintenance of the aircraft are all possible. In order to present a competitive financial model for sustained drone services, EcoSoar is made with reliable low-cost materials and electronics. This paper lays out the rapid prototyping and flight experiment efforts that went into polishing the design, test results from an EcoSoar centered drone workshop in Kasungu, Malawi, and finally a range optimization study with flight test validation. / Master of Science / A new humanitarian drone has been developed at the Unmanned Systems Lab (USL) at Virginia Tech. The unmanned aerial vehicle (UAV), named EcoSoar, was designed with the goal of creating a business opportunity for local entrepreneurs in low-resource communities. In order to be a viable solution in the developing world EcoSoar utilizes customizable 3D-printed parts and wings made from cheap materials like posterboard and packing tape. In addition, tools for building the drone have been developed in such a way that anyone can learn to construct and operate EcoSoar regardless of experience. This paper lays out the engineering efforts that went into the design, lessons learned from an EcoSoar-centered workshop in Kasungu, Malawi, and finally offers an upgraded design.

Skogsbränder är svåra att upptäcka i ett tidigt stadie, vilket leder till förödande konsekvenser. Hela 30 % av koldioxiden som atmosfären tar emot kommer från skogsbränder. Flera tusentals människor och djur mister livet eller tvingas lämna sina hem. Det finns idag flera tekniker som med varierande framgång kan upptäcka skogsbränder. I detta arbete skall en alternativ metod för rökdetektering utvecklas och testas. Metoden ska vara möjlig att appliceras på UAV (Unmanned Aerial Vehicle) teknik. Arbetet fokuserar på att skilja på brandrök och dimma med högupplöst data. Två algoritmer prövas, SDA (Statistisk distributions algoritm) och KBA (Kunskapsbaserad igenkännings algoritm). Den första testar statistiska distributioner för att hitta unika identifierare för rök. Den andra algoritmen är baserad på kunskapen om rök vad gäller spektrala och morfologiska egenskaper. Röken identifieras med hjälp av form, area och kanter. Algoritmen visade en precision med 90 % i bilder innehållande rök och en feldetektering med 20 % för bilder innehållande dimma. / It is very difficult to discover forest fires in an early stage which can lead to devastat-ing consequences. Today, 30% of the total carbon dioxide that is released in the at-mosphere comes from forest fires. Thousands of human beings and animals are killed or forced to leave their homes every year. There are a variety of techniques today that is being used for discovering forest fires but whom lack in accuracy or has problems with a large amount of false alarms. This paper is an experimental study to try to solve this issue. The proposed method in this paper could be applied on UAV (Unmanned Arial Vehicles). This study will focus on identifying smoke regions from forest fires and removing fog objects which has similar characteristics as smoke. Two algorithms are tested, SDA (Statistical distributions algorithm) and KBA (Knowledge-based identification algorithm). The SDA uses statistic distribution al-gorithm where smoke and fogs characteristics are identified. The second algorithm, KBA, is a knowledge-based algorithm, where the shape, area and edges of the smoke’s characteristics are applied. The algorithm showed a 90 % accuracy for find-ing smoke in images with a false alarm rate of 20 % in images of fog.