Background Tarkett AB Ronneby (Sweden) is a flooring solutions company, recognized for the manufacturing and recycling of homogeneous plastic flooring. Tarkett AB recycles mainly installation spill and manufacturing defects. However, Tarkett AB is considering widening its recycling capabilities to include old and torn plastic floors which may contain impurities and banned substances or plastic floors of competing brands. To accomplish this, Tarkett is considering a completely new recycling line with an automated sorting process instead of the current manual process. Thus, Tarkett proposes a dissertation to conceptualize a new automated sorting system with added capacity and increased functionality. Purpose This work aims to investigate the current sorting process and introduce conceptual solutions for a new automated sorting process capable of identifying and separating plastic floors according to the manufacturer, type, condition, and external waste by using existing technology. Method The methods and tools used in this work are mainly based on a modified product development process. Starting with data collection of the current sorting process, performing a need-finding, and extracting requirements for an automated sorting process, investigating relevant technology, evaluating technology based on scientific literature and tests. The testing was conducted in collaboration with two companies. Near-infrared scanners were tested with Holger AB, while pattern recognition systems were tested with Vision-Geek. Finally, three concepts for the automated sorting process were developed and shown through flow charts and 2D-3D illustrations. Results The results of this work showed that it was possible to use near-infrared and pattern recognition for the separation of plastic floors. Besides, three conceptual solutions for an automated sorting process were generated and showcased with schematic graphs and 2D-3D illustrations. The concepts describe how the sorting process functions and what technology is used for each step of the process. Concept 1 and Concept 2 used both pattern recognition and spectroscopy methods. While Concept 3 only used spectroscopy methods. Moreover, spectroscopy methods were used to sort plastic floors by content while pattern recognition by appearance. Conclusions Recycling of torn and old plastic flooring can be beneficial for both the environment and the recycling industry. Yet, it presents some challenges relating to reliable, fast, and nondestructive identification for sorting and separation purposes. New and proven technology such as near-infrared hyperspectral imaging and pattern recognition can be used. However, high-quality pattern and spectrum libraries of multiple plastic floors have to be created for optimal and reliable reference models. Furthermore, pattern recognition and near-infrared methods need to be tested further at an industrial scale. / Bakgrund Tarkett AB Ronneby (Sverige) är ett golvlösning företag, erkänt för tillverkning och återvinning av homogent plastgolv. Tarkett AB återvinner huvudsakligen installations spill och tillverkningsfel. Tarkett AB överväger dock att utvidga sina återvinnings förmågor till att omfatta gamla och sönderrivna plastgolv som kan innehålla föroreningar och förbjudna ämnen eller plastgolv från konkurrerande varumärken. För att åstadkomma detta överväger Tarkett en helt ny återvinnings linje med en automatiserad sorteringsprocess istället för den aktuella manuella processen. Således föreslår Tarkett ett examensarbete för att konceptualisera ett nytt automatiserat sorteringssystem med ökad kapacitet och ökad funktionalitet. Syfte Detta arbete syftar till att undersöka den nuvarande sorterings processen och introducera konceptuella lösningar för en ny automatiserad sorteringsprocess som kan identifiera och separera plastgolv efter tillverkare, typ, skick och externt avfall med befintlig teknik. Metod De metoder och verktyg som används i detta arbete är huvudsakligen baserade på en modifierad produktutvecklingsprocess. Vilket börja med datainsamling av den aktuella sorterings processen, hitta behov och extrahera krav för en automatiserad sorteringsprocess, undersöka relevant teknik, utvärdera tekniken baserad på vetenskaplig litteratur och tester. Testningen genomfördes i samarbete med två företag. Nära-infraröda skannrar testades med Holger AB, medan mönsterigenkänning system testades med Vision-Geek. Slutligen utvecklades tre koncept för den automatiserade sorterings processen och visades genom flödesscheman och 2D-3D-illustrationer. Resultat Resultaten av detta arbete visade att det var möjligt att använda nära-infraröd och mönsterigenkänning för separering av plastgolv. Dessutom genererades tre konceptuella lösningar för en automatiserad sorteringsprocess och visades med schematiska grafer och 2D-3D-illustrationer. Begreppen beskriver hur sorterings processen fungerar och vilken teknik som används för varje steg i processen. Koncept 1 och Koncept 2 använde både mönsterigenkänning och spektroskopi metoder. Medan Koncept 3 bara använde spektroskopi metoder. Spektroskopi metoderna användes för att sortera plastgolv efter innehåll medan mönsterigenkänning efter utseende. Slutsats Återvinning av sönderrivna plastgolv kan vara fördelaktigt för både miljön och återvinningsindustrin. Dock finns det några utmaningar med anknytning till pålitlig, snabb och icke-förstörande identifiering för sorterings- och separation ändamål. Ny och beprövad teknik som nästan infraröd hyperspektral avbildning och mönsterigenkänning kan användas. Emellertid måste mönster- och spektrum bibliotek av hög kvalitet av flera plastgolv skapas för optimala och pålitliga referens-modeller. Dessutom måste mönsterigenkänning och nära-infraröda metoder testas vidare i industriell skala.

The population of northern pike in the Baltic sea has seen a drasticdecrease in numbers in the last couple of decades. The reasons for this are believed to be many, but the majority of them are most likely anthropogenic. Today, many measures are being taken to prevent further decline of pike populations, ranging from nutrient runoff control to habitat restoration. This inevitably gives rise to the problem addressed in this project, namely: how can we best monitor pike populations so that it is possible to accurately assess and verify the effects of these measures over the coming decades? Pike is currently monitored in Sweden by employing expensive and ineffective manual methods of individual marking of pike by a handful of experts. This project provides evidence that such methods could be replaced by a Convolutional Neural Network (CNN), an automatic artificial intelligence system, which can be taught how to identify pike individuals based on their unique patterns. A neural net simulates the functions of neurons in the human brain, which allows it to perform a range of tasks, while a CNN is a neural net specialized for this type of visual recognition task. The results show that the CNN trained in this project can identify pike individuals in the provided data set with upwards of 90% accuracy, with much potential for improvement.

In the forecast data post-processing at the Swedish Meteorological and Hydrological Institute (SMHI) a regular Kalman filter is used to debias the two meter air temperature forecast of the physical models by controlling towards air temperature observations. The Kalman filter however diverges when encountering greater nonlinearities in shifting weather patterns, and can only be manually reset when a new air mass has stabilized itself within its operating region. This project aimed to automate this process by means of a machine learning approach. The methodology was at its base supervised learning, by first algorithmically labelling the air mass shift occurrences in the data, followed by training a logistic regression model. Observational data from the latest twenty years of the Uppsala automatic meteorological station was used for the analysis. A simple pipeline for loading, labelling, training on and visualizing the data was built. As a work in progress the operating regime was more of a semi-supervised one - which also in the long run could be a necessary and fruitful strategy. Conclusively the logistic regression appeared to be quite able to handle and infer from the dynamics of air temperatures - albeit non-robustly tested - being able to correctly classify 77% of the labelled data. This work was presented at Uppsala University in June 1st of 2018, and later in June 20th at SMHI.

Vector Quantization (VQ) is a classic optimization problem and a simple approach to pattern recognition. Applications include lossy data compression, clustering and speech and speaker recognition. Although VQ has largely been replaced by time-aware techniques like Hidden Markov Models (HMMs) and Dynamic Time Warping (DTW) in some applications, such as speech and speaker recognition, VQ still retains some significance due to its much lower computational cost — especially for embedded systems. A recent study also demonstrates a multi-section VQ system which achieves performance rivaling that of DTW in an application to handwritten signature recognition, at a much lower computational cost. Adding sensitivity to temporal patterns to a VQ algorithm could help improve such results further. SOTPAR2 is such an extension of Neural Gas, an Artificial Neural Network algorithm for VQ. SOTPAR2 uses a conceptually simple approach, based on adding lateral connections between network nodes and creating “temporal activity” that diffuses through adjacent nodes. The activity in turn makes the nearest-neighbor classifier biased toward network nodes with high activity, and the SOTPAR2 authors report improvements over Neural Gas in an application to time series prediction. This report presents an investigation of how this same extension affects quantization and prediction performance of the self-organizing incremental neural network (SOINN) algorithm. SOINN is a VQ algorithm which automatically chooses a suitable codebook size and can also be used for clustering with arbitrary cluster shapes. This extension is found to not improve the performance of SOINN, in fact it makes performance worse in all experiments attempted. A discussion of this result is provided, along with a discussion of the impact of the algorithm parameters, and possible future work to improve the results is suggested. / Vektorkvantisering (VQ; eng: Vector Quantization) är ett klassiskt problem och en enkel metod för mönsterigenkänning. Bland tillämpningar finns förstörande datakompression, klustring och igenkänning av tal och talare. Även om VQ i stort har ersatts av tidsmedvetna tekniker såsom dolda Markovmodeller (HMM, eng: Hidden Markov Models) och dynamisk tidskrökning (DTW, eng: Dynamic Time Warping) i vissa tillämpningar, som tal- och talarigenkänning, har VQ ännu viss relevans tack vare sin mycket lägre beräkningsmässiga kostnad — särskilt för exempelvis inbyggda system. En ny studie demonstrerar också ett VQ-system med flera sektioner som åstadkommer prestanda i klass med DTW i en tillämpning på igenkänning av handskrivna signaturer, men till en mycket lägre beräkningsmässig kostnad. Att dra nytta av temporala mönster i en VQ-algoritm skulle kunna hjälpa till att förbättra sådana resultat ytterligare. SOTPAR2 är en sådan utökning av Neural Gas, en artificiell neural nätverk-algorithm för VQ. SOTPAR2 använder en konceptuellt enkel idé, baserad på att lägga till sidleds anslutningar mellan nätverksnoder och skapa “temporal aktivitet” som diffunderar genom anslutna noder. Aktiviteten gör sedan så att närmaste-granne-klassificeraren föredrar noder med hög aktivitet, och författarna till SOTPAR2 rapporterar förbättrade resultat jämfört med Neural Gas i en tillämpning på förutsägning av en tidsserie. I denna rapport undersöks hur samma utökning påverkar kvantiserings- och förutsägningsprestanda hos algoritmen självorganiserande inkrementellt neuralt nätverk (SOINN, eng: self-organizing incremental neural network). SOINN är en VQ-algorithm som automatiskt väljer en lämplig kodboksstorlek och också kan användas för klustring med godtyckliga klusterformer. Experimentella resultat visar att denna utökning inte förbättrar prestandan hos SOINN, istället försämrades prestandan i alla experiment som genomfördes. Detta resultat diskuteras, liksom inverkan av parametervärden på prestandan, och möjligt framtida arbete för att förbättra resultaten föreslås.

Convolutional artificial neural networks can be applied for image-based object classification to inform automated actions, such as handling of objects on a production line. The present thesis describes theoretical background for creating a classifier and explores the effects of introducing a set of relatively recent techniques to an existing ensemble of classifiers in use for an industrial sorting system.The findings indicate that it's important to use spatial variety dropout regularization for high resolution image inputs, and use an optimizer configuration with good convergence properties. The findings also demonstrate examples of ensemble classifiers being effectively consolidated into unified models using the distillation technique. An analogue arrangement with optimization against multiple output targets, incorporating additional information, showed accuracy gains comparable to ensembling. For use of the classifier on test data with statistics different than those of the dataset, results indicate that augmentation of the input data during classifier creation helps performance, but would, in the current case, likely need to be guided by information about the distribution shift to have sufficiently positive impact to enable a practical application. I suggest, for future development, updated architectures, automated hyperparameter search and leveraging the bountiful unlabeled data potentially available from production lines.