Orientador : Maria Cecilia Sanchez Teixeira / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Educação / Made available in DSpace on 2018-07-27T16:13:34Z (GMT). No. of bitstreams: 1 Wrege_RachelSilveira_D.pdf: 1336242 bytes, checksum: b03fbdb01fca95492b424a262a1b8bd9 (MD5) Previous issue date: 2001 / Doutorado

Teknikens utveckling går i en rasande fart och idag kan man köpa en avancerad obemannad luftfarkost i butik. De obemannade luftfarkosterna kallas för multirotorfarkoster, eller i folkmun för drönare. Dessa multirotorfarkoster kan leverera en dödlig last med stor precision av någon utan större tekniskt kunnande. I stadsmiljö vid stora folkansamlingar skulle det idag vara svårt att stoppa en sådan inkommande attack utan att riskera ytterligare skador. Det här arbetet går ut på att belysa problemet och studera tekniken. Eftersom tekniken bara blir mer och mer avancerad så gäller det att vara vaksam för att förebygga en terrorattack. Ibland kanske det inte räcker att vara vaksam och då måste man söka andra lösningar. Från studien har ett antal krav på en försvarande lösning sammanställts. Från de utarbetade kraven så har också ett koncept tagits fram i form av en försvarande multirotor. / The developments in technology are at a furious pace and today you can buy an advanced unmanned aerial vehicle at hobby retail stores. These unmanned aerial vehicles are called multi-rotorcrafts or in the vernacular for drones. These multi-rotorcrafts can deliver a lethal payload with great precision by someone without much technical knowledge. In the urban environment at large public gatherings, it is currently difficult to stop an incoming attack without risking collateral damage. This papers purpose is to highlight the problem and study the technology. As the technology is just becoming more and more advanced you have to be vigilant to prevent a terrorist attack. It may not be enough and then you have to look for other solutions. From the study a number of requirements for a defending solution have been compiled. Also a concept has been developed from the elaborated requirements in shape of a defending multi-rotorcraft.

I examensarbetet jämförs undvikande av kollision och tidsefektivitet mellan det två metoderna hastighetshinder och partikelseparation i spelgenren Twin stick shooter. Arbetet försöker besvara frågan: Hur skiljer sig undvikandet av kollision och tidseffektiviteten mellan metoderna hastighetshinder och partikelseparation, i spelgenren twin stick shooter med flockbeteende? För att besvara frågan har en artefakt skapats. I artefakten jagar agenter en spelare medan agenterna undviker kollision med andra agenter, dock eftersträvar agenterna att kollidera med spelaren. I artefakten körs olika experiment baserat på parametrar som har ställts in. Varje experiment körs en bestämd tid och all data om kollisioner och exekveringstid för respektive metod sparas i en textfil.   Resultatet av experimenten pekar på att partikelseparation lämpar sig bättre för twin stick shooters.  Hastighetshinder kolliderar mindre men tidsberäkningen är för hög och skalar dåligt med antal agenter. Det passar inte twinstick shooter då det oftast är många agenter på skärmen.  Metoderna för undvikandet av kollision har användning till radiostyrda billar och robotar, samt simulation av folkmassa.

Fokus i examensarbetet har varit att konstruera och programmera en självstyrande bil med avgränsningar till regelverket i tävlingen Freescale cup. Då studien befinner sig i en tävlingsdomän, blir precision och hastighet två huvudfaktorer för linjeföljning och navigering. För att bemöta de ovanstående problem, fokuseras studien i denna rapport på utvecklandet av en ljusbehandlar-algoritm. Detta för hantera och standardisera linjeavkodningen och linjeidentifieringen. Detta kombinerat med en primärreglering som hanterar styrningen med avseende på linjeidentifieringen med tidsförskjutning för sekundärreglering vid behov. För att mäta precision utfördes tester genom tre olika metoder på linjeavläsningen: (1) avläsning av sensordata utan filtrering,(2) avläsning av sensordata via Kalmanfilter samt (3) avläsning av sensordata med en adaptiv ljusalgoritm för de olika fälten och ljusförhållanden. Dessa i kombination med väl injusterad PID-styralgoritm borgade för god precision och linjeföljning under dåliga förhållanden.   Studien påvisar att med den adaptiva ljusalgoritmen i kombination med linjeföljaralgoritmen och sekundärreglering via look ahead-algoritmen kan hastigheter som tangerar optimala förhållanden även uppnås under mer kaotiska och dåliga förhållanden. Bilen finner sin väg trots ljusförhållanden så låga som 20 lux. Studien styrker även fördelarna med den adaptiva ljusalgoritmen i kombination med linjeföljaralgoritmen och look ahead-algoritmen som primärreglering under optimala driftförhållanden.   Effektiviteten av studien och de algoritmer de mynnades ut i kvantifierades och jämfördes genom varvtider och dess linjeföljarförmåga under olika ljusförhållanden och olika gynnsamma underlag. Som ytterligare ett led för att få upp snitthastigheten med aktiv chassidynamik, nyttjades styrvinkeln som en funktion till differentialdriften av motorerna för att bibehålla sitt väggrepp genom kurvorna och hastigheten med aktiv hastighetskontroll.

In this thesis, the development of a stereo camera system for an intelligent tool is presented. The task of the system is to identify and localise objects so that the tool can guide a robot. Different approaches to object detection have been implemented and evaluated and the systems ability to localise objects has been tested. The results show that the system can achieve a localisation accuracy below 5 mm.

Thermal spectrum cameras are gaining interest in many applications due to their long wavelength which allows them to operate under low light and harsh weather conditions. One disadvantage of thermal cameras is their limited visual interpretability for humans, which limits the scope of their applications. In this thesis, we try to address this problem by investigating the possibility of transforming thermal infrared (TIR) images to perceptually realistic visible spectrum (VIS) images by using Convolutional Neural Networks (CNNs). Existing state-of-the-art colorization CNNs fail to provide the desired output as they were trained to map grayscale VIS images to color VIS images. Instead, we utilize an auto-encoder architecture to perform cross-spectral transformation between TIR and VIS images. This architecture was shown to quantitatively perform very well on the problem while producing perceptually realistic images. We show that the quantitative differences are insignificant when training this architecture using different color spaces, while there exist clear qualitative differences depending on the choice of color space. Finally, we found that a CNN trained from day time examples generalizes well on tests from night time.

Training data is the bottleneck for training Convolutional Neural Networks. A larger dataset gives better accuracy though also needs longer training time. It is shown by finetuning neural networks on synthetic rendered images, that the mean average precision increases. This method was applied to two different datasets with five distinctive objects in each. The first dataset consisted of random objects with different geometric shapes. The second dataset contained objects used to assemble IKEA furniture. The neural network with the best performance, trained on 5400 images, achieved a mean average precision of 0.81 on a test which was a sample of a video sequence. Analysis of the impact of the factors dataset size, batch size, and numbers of epochs used in training and different network architectures were done. Using synthetic images to train CNN’s is a promising path to take for object detection where access to large amount of annotated image data is hard to come by.

In a synchronized multi camera system it is imperative that the synchronization error between the different cameras is as close to zero as possible and the jitter of the presumed frame rate is as small as possible. It is even more important when these systems are used in an autonomous vehicle trying to sense its surroundings. We would never hand over the control to a autonomous vehicle if we couldn't trust the data it is using for moving around. The purpose of this thesis was to build a synchronization setup for a multi camera system using state of the art RayTrix digital cameras that will be used in the iQMatic project involving autonomous heavy duty vehicles. The iQMatic project is a collaboration between several Swedish industrial partners and universities. There was also software development for the multi camera system involved. Different synchronization techniques were implemented and then analysed against the system requirements. The two techniques were hardware trigger i.e. external trigger using a microcontroller, and software trigger using the API from the digital cameras. Experiments were conducted by testing the different trigger modes with the developed multi camera software. The conclusions show that the hardware trigger is preferable in this particular system by showing more stability and better statistics against the system requirements than the software trigger. But the thesis also show that additional experiments are needed for a more accurate analysis. / iQMatic

Thermal spectrum cameras are gaining interest in many applications due to their long wavelength which allows them to operate under low light and harsh weather conditions. One disadvantage of thermal cameras is their limited visual interpretability for humans, which limits the scope of their applications. In this thesis, we try to address this problem by investigating the possibility of transforming thermal infrared (TIR) images to perceptually realistic visible spectrum (VIS) images by using Convolutional Neural Networks (CNNs). Existing state-of-the-art colorization CNNs fail to provide the desired output as they were trained to map grayscale VIS images to color VIS images. Instead, we utilize an auto-encoder architecture to perform cross-spectral transformation between TIR and VIS images. This architecture was shown to quantitatively perform very well on the problem while producing perceptually realistic images. We show that the quantitative differences are insignificant when training this architecture using different color spaces, while there exist clear qualitative differences depending on the choice of color space. Finally, we found that a CNN trained from daytime examples generalizes well on tests from night time.

This master thesis explores the possibility of using generative Adversarial Networks (GANs) to refine labeled synthetic code images to resemble real code images while preserving label information. The GAN used in this thesis consists of a refiner and a discriminator. The discriminator tries to distinguish between real images and refined synthetic images. The refiner tries to fool the discriminator by producing refined synthetic images such that the discriminator classify them as real. By updating these two networks iteratively, the idea is that they will push each other to get better, resulting in refined synthetic images with real image characteristics. The aspiration, if the exploration of GANs turns out successful, is to be able to use refined synthetic images as training data in Semantic Segmentation (SS) tasks and thereby eliminate the laborious task of gathering and labeling real data. Starting off from a foundational GAN-model, different network architectures, hyperparameters and other design choices are explored to find the best performing GAN-model. As is widely acknowledged in the relevant literature, GANs can be difficult to train and the results in this thesis are varying and sometimes ambiguous. Based on the results from this study, the best performing models do however perform better in SS tasks than the unrefined synthetic set they are based on and benchmarked against, with regards to Intersection over Union.