Semantic segmentation of a scene aims to give meaning to the scene by dividing it into meaningful — semantic — parts. Understanding the scene is of great interest for all kinds of autonomous systems, but manual annotation is simply too time consuming, which is why there is a need for an alternative approach. This thesis investigates the possibility of automatically segmenting 3D-models of urban scenes, such as buildings, into a predetermined set of labels. The approach was to first acquire ground truth data by manually annotating five 3D-models of different urban scenes. The next step was to extract features from the 3D-models and evaluate which ones constitutes a suitable feature space. Finally, three supervised learners were implemented and evaluated: k-Nearest Neighbour (KNN), Support Vector Machine (SVM) and Random Classification Forest (RCF). The classifications were done point-wise, classifying each 3D-point in the dense point cloud belonging to the model being classified. The result showed that the best suitable feature space is not necessarily the one containing all features. The KNN classifier got the highest average accuracy overall models — classifying 42.5% of the 3D points correct. The RCF classifier managed to classify 66.7% points correct in one of the models, but had worse performance for the rest of the models and thus resulting in a lower average accuracy compared to KNN. In general, KNN, SVM, and RCF seemed to have different benefits and drawbacks. KNN is simple and intuitive but by far the slowest classifier when dealing with a large set of training data. SVM and RCF are both fast but difficult to tune as there are more parameters to adjust. Whether the reason for obtaining the relatively low highest accuracy was due to the lack of ground truth training data, unbalanced validation models, or the capacity of the learners, was never investigated due to a limited time span. However, this ought to be investigated in future studies.

No description available.

This work investigates the landscape of aerial image stereo matching (AISM) methods suitable for large scale forest variable estimation. AISM methods are an important source of remotely collected information used in modern forestry to keep track of a growing forest's condition. A total of 17 AISM methods are investigated, out of which 4 are evaluated by processing a test data set consisting of three aerial images. The test area is located in southern Sweden, consisting of mainly Norway Spruce and Scots Pine. From the resulting point clouds and height raster images, a total of 30 different metrics of both height and density types are derived. Linear regression is used to fit functions from metrics derived from AISM data to a set of forest variables including tree height (HBW), tree diameter (DBW), basal area, volume. As ground truth, data collected by dense airborne laser scanning is used. Results are presented as RMSE and standard deviation concluded from the linear regression. For tree height, tree diameter, basal area, volume the RMSE ranged from 7.442% to 10.11%, 11.58% to 13.96%, 32.01% to 35.10% and 34.01% to 38.26% respectively. The results concluded that all four tested methods achieved comparable estimation quality although showing small differences among them. Keystone and SURE performed somewhat better while MicMac placed third and Photoscan achieved the less accurate result.

Runway detection in long wavelength infrared (LWIR) video could potentially increase the number of successful landings by increasing the situational awareness of pilots and verifying a correct approach. A method for detecting runways in LWIR video was therefore proposed and evaluated for robustness, speed and FPGA acceleration. The proposed algorithm improves the detection probability by making assumptions of the runway appearance during approach, as well as by using a modified Hough line transform and a symmetric search of peaks in the accumulator that is returned by the Hough line transform. A video chain was implemented on a Xilinx ZC702 Development card with input and output via HDMI through an expansion card. The video frames were buffered to RAM, and the detection algorithm ran on the CPU, which however did not meet the real-time requirement. Strategies were proposed that would improve the processing speed by either acceleration in hardware or algorithmic changes.

The usage of 3D modelling is increasing fast, both for civilian and military areas, such as navigation, targeting and urban planning. When creating a 3D model from satellite images, clouds canbe problematic. Thus, automatic detection ofclouds inthe imagesis ofgreat use. This master thesis was carried out at Vricon, who produces 3D models of the earth from satellite images.This thesis aimed to investigate if Support Vector Machines could classify pixels into cloud or non-cloud, with a combination of texture and color as features. To solve the stated goal, the task was divided into several subproblems, where the first part was to extract features from the images. Then the images were preprocessed before fed to the classifier. After that, the classifier was trained, and finally evaluated.The two methods that gave the best results in this thesis had approximately 95 % correctly classified pixels. This result is better than the existing cloud segmentation method at Vricon, for the tested terrain and cloud types.

Examensarbetet bakom denna rapport utgör en undersökande studie om hur transmission av mediadata i nätverk kan göras effektivare. Det kan åstadkommas genom att viss tilläggsinformation avsedd för att jämna ut datatakten adderas i det realtidsprotokoll, Real Time Protocol, som används för strömmande media. Genom att försöka skicka lika mycket data under alla konsekutiva tidsintervall i sessionen kommer datatakten vid en godtycklig tidpunkt med större sannolikhet att vara densamma som vid tidigare klockslag. En streamingserver kan tolka, hantera och skicka data vidare enligt instruktionerna i protokollets sidhuvud. Datatakten jämnas ut genom att i förtid, under tidsintervall som innehåller mindre data, skicka även senare data i strömmen. Resultatet av detta är en utjämnad datataktskurva som i sin tur leder till en jämnare användning av nätverkskapaciteten. Arbetet inkluderar en översiktlig analys av beteendet hos strömmande media, bakgrundsteori om filkonstruktion och nätverksteknologier samt ett förslag på hur mediafiler kan modifieras för att uppfylla syftet med examensarbetet. Resultat och diskussion kan förhoppningsvis användas som underlag för en framtida implementation av en applikation ämnad att förbättra trafikflöden över nätverk.

Mine automation has been in development since the 1980s and began to be implemented in the 1990s with the biggest drivers being safety, reduced maintenance and increased productivity. Automation is used in many different mining methods performing a variety of tasks. However, very few studies have been conducted regarding the performance of automatic vehicles in real world mines, neither has much research been done when comparing automated and manual loading.   The purpose of this thesis is twofold. First it is to identify and produce metrics that can be used to compare automatic and manual loading in an underground mining environment. A literary study is presented as a basis for these metrics where the development history is chronicled and the mechanics by which an automated system controls an automatic vehicle is explained. Also included is a description of different mining methods and the part that automation plays in them.   Secondly, the goal is to use these metrics in a real world case study of automated loading in an actual operational mine. The Garpenberg mine, owned by Boliden Minerals, is an underground metal mine located in Dalarna, Sweden, and mining has been done there since the 1200s. Load-Haul-Dump (LHD) machines are used to load and haul ore ether using automation or tele operation. For this study information is extracted from Bolidens internal databases to calculate the metrics which are then used to analyze the performance and reliability of automated loaders, and also to compare manual and automatic loading.   There are mainly three types of metrics that are relevant to automatic and manual loading, these being time related metrics, production related metrics and reliability related metrics.   Since the LHDs dealt with in this thesis are operated both manually and automatically the main use of the time related metrics is to find the relationship between modes of operation in regards to engine hours spent in action, the amount and character of downtime that occurred during the analyzed time span, and the utilization of automatic LHDs during the workday. The most straightforward comparison between manual and automatic loading is the production, i.e. the tonnage of loaded material per unit time. In this thesis production will be analyzed per hour and per loading activity (which run between 3 and 4 hours). Lastly automated and manual loading will be compared on the basis of availability and maintenance, the reason being that LHDs are driven in different ways depending on the mode of operation.   The results show that automatic loading consists of between 14 and 29% of production time while contributing to between 17 and 28% of downtime. No clear connection can be seen between downtime and the relationship of automatic to manual loading, as the difference is not bigger than 10% either way. Manual loading constitutes between 2500 and 3250 engine hours, while time spent during automatic loading constitutes between 420 and 1095 engine hours. The majority of all downtime is not specific to either mode of operation, although automatic specific stops constitutes up to 50% of total downtime for two of the LHDs studied. The distribution of loading activities is similar for both manual and automatic loading, with the number of concurrent activities dipping during lunch breaks and stopping entirely during shift changes. Manual loading peaks around 1000 concurrent jobs while automatic peaks at just fewer than 200.   Regarding production the results show that manual loading is more effective in the short term, as three out of four LHDs has had a higher tonnage loaded per hour. However, when locking at the tonnage loaded per activity, automatic loading catches up to manual loading and produces more tonnage in the long term, with three out of four LHDs getting this higher production.   One interesting result can be found in regards to fuel economy, as one of the LHDs show a reduced consumption of fuel while at the same time having had the largest percentage of time spent in automatic mode. No clear connection can be seen with the rest of the LHDs however, as they show no clear connection between time spent in ether mode of operation and the amount of fuel consumed.   Automatic loading proved to have slightly higher availability than manual loading. In all cases however, the difference in no more than 10% and both modes of operation is above 90% availability. The higher availability of automatic loading is attributed to the fact that manual loading constitutes much more time than automatic loading, and thus there has been more time for breakdowns and production stops to occur for manual loading. The relationship of preventive and corrective maintenance is the same for all four LHDs irrespective of amount of time spent in ether operating mode. Preventive maintenance jobs accounts for more than 90% of the number maintenance actions for all LHDs.   When analyzing what kinds of production stops are the most prevalent, there are differences between manual and automatic loading. For manual loading the most common stops are those that have to do with external circumstances in the mine such as blocked access and fallen boulders, and those to do with minor breakdowns of the LHD. For automatic loading the most common stops are those to do with the automatic system and the equipment used to operate the automated LHD, followed by those caused by external circumstances, similarly to manual loading. Automatic loading has proportionally fewer stops than manual loading in all categories except those unique to automation, which is in turn the biggest category of all production stops.   The conclusions that can be drawn from these results are that automatic loading can outperform manual loading in the long term, but that continuous uninterrupted loading activates are important to achieve this. Automatic and manual loading show comparable reliability when it comes to maintenance and repairs (serious breakdowns are very rare). Availability and the relationship of preventive and corrective maintenance are similar between both modes of operation. The analysis of production stops show that the biggest problem with automatic loading is the automatic systems and the specialized equipments inability to handle the underground environment. Problems with recorded routes and falsely tripped safety systems are the most common stops.       Recommendations to Boliden Minerals regarding the automatic system consist of improving remote troubleshooting and streamlining of problem solving dealing with automation software and hardware. Steps should also be takes towards tailoring the underground environment to better suit automation.   Suggestions to further research consist of deeper studies of all the metrics presented in this thesis to better analyze the role of automation in the global mining industry. Another avenue of study is the combination of the findings in this thesis with the actual environment and layout in the Garpenberg mine to better understand the connection between operating environment and the reliability and productivity of the automatic system.

In this thesis spacetime analysis is applied to laser triangulation in an attempt to eliminate certain artifacts caused mainly by reflectance variations of the surface being measured. It is shown that spacetime analysis do eliminate these artifacts almost completely, it is also shown that the shape of the laser beam used no longer is critical thanks to the spacetime analysis, and that in some cases the laser probably even could be exchanged for a non-coherent light source. Furthermore experiments of running the derived algorithm on a GPU (Graphics Processing Unit) are conducted with very promising results. The thesis starts by deriving the theory needed for doing spacetime analysis in a laser triangulation setup taking perspective distortions into account, then several experiments evaluating the method is conducted.

Detta är en reflexiv rapport om arbetsmetoden bakom skapandet av en filmtrailer som ska marknadsföra den fiktiva filmen om Gabriel Glömmer baserad på en bok av Ulf Löfgren. Boken handlar om en tillbakadragen pojke med en livlig fantasi som drömmer sig till en sagovärld där han får uppleva äventyr och möta nya vänner. Texten har ett fokus på skapandet av miljöer i 3D, med mål att behålla den grafiska stilen från 2D illustrationer som finns i boken. Resultatet av projektet är en filmtrailer med längden 1 minut och 40 sekunder. Filmtrailerns bilder har en uppbyggnad som försöker efterlikna de illustrationer som finns i boken.

Rapporten tar upp mina reflektioner dels angående projektarbetet både på ett personligt plan samt samarbetet mellan oss två som gjorde projektet. Vad har fungerat och vad har fungerat mindre bra under projekttiden? Från 2D till 3D är en röd tråd genom hela rapporten då vi gått från 2D-referernsbilder i barnboken Gabriel Glömmer till utrenderade 3D-sekvenser. Jag går igenom vad jag gjort och vad andra gjort under projekttiden då vi haft hjälp av lite olika människor. Dessa människor presenteras snabbt, något annat som presenteras är boken och min relation till densamma. Resultatet av projektarbetet blev att vi lyckades med de uppställde mål vi hade innan projektet startades.