No description available.

A mesura que la tecnologia evoluciona, apareixen noves aplicacions en el mon de la classificació facial. En el reconeixement de patrons, normalment veiem les cares com a punts en un espai de alta dimensionalitat definit pels valors dels seus pixels. Aquesta aproximació pateix diversos problemes: el fenomen de la "la maledicció de la dimensionalitat", la presència d'oclusions parcials o canvis locals en la il·luminació. Tradicionalment, només les característiques internes de les imatges facials s'han utilitzat per a classificar, on normalment es fa una extracció de característiques. Les tècniques d'extracció de característiques permeten reduir la influencia dels problemes mencionats, reduint també el soroll inherent de les imatges naturals alhora que es poden aprendre característiques invariants de les imatges facials. En la primera part d'aquesta tesi presentem alguns mètodes d'extracció de característiques clàssics: Anàlisi de Components Principals (PCA), Anàlisi de Components Independents (ICA), Factorització No Negativa de Matrius (NMF), i l'Anàlisi Discriminant de Fisher (FLD), totes elles fent alguna mena d'assumpció en les dades a classificar. La principal contribució d'aquest treball es una nova família de tècniques d'extracció de característiques usant el algorisme del Adaboost. El nostre mètode no fa cap assumpció en les dades a classificar, i construeix de forma incremental la matriu de projecció tenint en compte els exemples mes difícilsPer altra banda, en la segon apart de la tesi explorem el rol de les característiques externes en el procés de classificació facial, i presentem un nou mètode per extreure un conjunt alineat de característiques a partir de la informació externa que poden ser combinades amb les tècniques clàssiques millorant els resultats globals de classificació. / As technology evolves, new applications dealing with face classification appear. In pattern recognition, faces are usually seen as points in a high dimensional spaces defined by their pixel values. This approach must deal with several problems such as: the curse of dimensionality, the presence of partial occlusions or local changes in the illumination. Traditionally, only the internal features of face images have been used for classification purposes, where usually a feature extraction step is performed. Feature extraction techniques allow to reduce the influence of the problems mentioned, reducing also the noise inherent from natural images and learning invariant characteristics from face images. In the first part of this thesis some internal feature extraction methods are presented: Principal Component Analysis (PCA), Independent Component Analysis (ICA), Non Negative Matrix Factorization (NMF), and Fisher Linear Discriminant Analysis (FLD), all of them making some kind of the assumption on the data to classify. The main contribution of our work is a non parametric feature extraction family of techniques using the Adaboost algorithm. Our method makes no assumptions on the data to classify, and incrementally builds the projection matrix taking into account the most difficult samples.On the other hand, in the second part of this thesis we also explore the role of external features in face classification purposes, and present a method for extracting an aligned feature set from external face information that can be combined with the classic internal features improving the global performance of the face classification task.

Feature interactions occur when one feature interferes with the intended operation of another feature. To detect such interactions, each new feature must be tested against existing features. The detected interactions must then be resolved; many existing approaches to resolving interactions require the feature set be prioritized. Unfortunately, the cost to determine a priority ordering for a feature set increases dramatically as the number of features increases. This thesis explores strategies to decrease the cost of prioritizing features, and thus facilitates priority-based solutions to resolving feature interactions. Specifically, this thesis introduces a categorization approach that reduces the complexity of determining priorities for a large set of features by decomposing the prioritization problem. Our categorization approach reduces this cost by using abstraction to divide the system's features into categories based on their main goal or functionality (e.g., block unwanted calls, present call information). Next, in order to detect and resolve the interactions that occur between these seemingly unrelated categories, we identify a set of principles for proper system behaviour that define acceptable behaviour in the global system. For example, a call that should be blocked by a call-screening feature should never result in a voice connection. The categories are then ordered, such that adherence to the principles is optimized. We show that using category priorities, to order a large feature set, correctly resolves interactions between individual features and significantly reduces the cost to determine priority orderings. The four significant contributions that this thesis makes are: 1) the categorization of features, 2) the principles of proper system behaviour, 3) automatic generation of priority orderings for categories, and 4) devising several optimizations that reduce the search space when exploring call simulations during the automatic generation of the priority orderings. These contributions are examined with respect to the telephony domain and result in the identification of 12 feature categories and 9 principles of proper system behaviour. A Prolog model was also created to run call simulations on the categories, using the identified principles as correctness criteria. Our case studies showed the reduced cost of our categorization approach is approximately 1/10^(55) % of the cost of a traditional approach. Given this significant reduction in the cost and the ability of our model to accurately reproduce the manually identified priority orderings, we can confidently argue that our categorization approach was successful. The three main limitations of our categorization approach are: 1) not all features (e.g., 911 features in telephony) can be categorized or some categories will contain a small number of features, 2) the generated priority ordering may still need to be analyzed by a human expert, and 3) the run time for our automatic generation of priority orderings remains factorial with respect to the size of the number of categories. However, these limitations are small in comparison to the savings generated by the categorization approach.

No description available.

Den här rapporten har utrett möjligheten att automatiskt identifiera diken frånflygburet insamlat LiDAR-data. Den metod för identifiering som har valts harförst skapat en höjdbild från LiDAR-data. Därefter har den tagit fram kandidatertill diken genom att vektorisera resultatet från en linjedetektering. Egenskaper-na för dikeskandidaterna har sedan beräknats genom en analys av höjdprofilerför varje enskild kandidat, där höjdprofilerna skapats utifrån ursprungliga data.Genom att filtrera kandidaterna efter deras egenskaper kan dikeskartor med an-vändarspecificerade mått på diken presenteras i ett vektorformat som underlättarvidare användning. Rapporten beskriver hur algoritmen har implementerats ochpresenterar också exempel på resultat. Efter en analys av algoritmen samt förslagpå förbättringar presenteras den viktigaste behållningen av rapporten; Att det ärmöjligt med automatisk detektering av diken. / This Master’s thesis is investigating the possibility of automatically identifyingditches in airborne collected LiDAR data. The chosen approach to identificationcommences by creating an elevation picture from the LiDAR data. Then it usesthe result of a line detection to exhibit candidates for ditches. The properties forthe various candidates are calculated through an analysis of the elevation profile forthe candidates, where the elevation profiles are created from the original data. Byfiltering the candidates according to their calculated properties, maps with ditchesconforming to user-specified limits are created and presented in vector format.This thesis describes how the algorithm is implemented and gives examples ofresults. After an analysis of the algorithm and a proposal for improvements, itis suggested that automatic detection of ditches in LiDAR collected data is anachievable objective.

Feature interactions occur when one feature interferes with the intended operation of another feature. To detect such interactions, each new feature must be tested against existing features. The detected interactions must then be resolved; many existing approaches to resolving interactions require the feature set be prioritized. Unfortunately, the cost to determine a priority ordering for a feature set increases dramatically as the number of features increases. This thesis explores strategies to decrease the cost of prioritizing features, and thus facilitates priority-based solutions to resolving feature interactions. Specifically, this thesis introduces a categorization approach that reduces the complexity of determining priorities for a large set of features by decomposing the prioritization problem. Our categorization approach reduces this cost by using abstraction to divide the system's features into categories based on their main goal or functionality (e.g., block unwanted calls, present call information). Next, in order to detect and resolve the interactions that occur between these seemingly unrelated categories, we identify a set of principles for proper system behaviour that define acceptable behaviour in the global system. For example, a call that should be blocked by a call-screening feature should never result in a voice connection. The categories are then ordered, such that adherence to the principles is optimized. We show that using category priorities, to order a large feature set, correctly resolves interactions between individual features and significantly reduces the cost to determine priority orderings. The four significant contributions that this thesis makes are: 1) the categorization of features, 2) the principles of proper system behaviour, 3) automatic generation of priority orderings for categories, and 4) devising several optimizations that reduce the search space when exploring call simulations during the automatic generation of the priority orderings. These contributions are examined with respect to the telephony domain and result in the identification of 12 feature categories and 9 principles of proper system behaviour. A Prolog model was also created to run call simulations on the categories, using the identified principles as correctness criteria. Our case studies showed the reduced cost of our categorization approach is approximately 1/10^(55) % of the cost of a traditional approach. Given this significant reduction in the cost and the ability of our model to accurately reproduce the manually identified priority orderings, we can confidently argue that our categorization approach was successful. The three main limitations of our categorization approach are: 1) not all features (e.g., 911 features in telephony) can be categorized or some categories will contain a small number of features, 2) the generated priority ordering may still need to be analyzed by a human expert, and 3) the run time for our automatic generation of priority orderings remains factorial with respect to the size of the number of categories. However, these limitations are small in comparison to the savings generated by the categorization approach.

Fall among the elderly is an important health issue. Fall detection and movement tracking techniques are therefore instrumental in dealing with this issue. This thesis responds to the challenge of classifying different movement types as a part of a system designed to fulfill the need for a wearable device to collect data for fall and near-fall analysis. Four different fall activities (forward, backward, left and right), three normal activities (standing, walking and lying down) and near-fall situations are identified and detected. Different machine learning algorithms are compared and the best one is used for the real time classification. The comparison is made using Waikato Environment for Knowledge Analysis or in short WEKA. The system also has the ability to adapt to different gaits of different people. A feature selection algorithm is also introduced to reduce the number of features required for the classification problem.

An experimental approach to traffic flow analysis is presented in which methodology from pattern recognition is applied to a specific dataset to examine its utility in determining traffic patterns. The selected dataset for this work, taken from a 1985 study by JHK and Associates (traffic research) for the Federal Highway Administration, covers an hour long time period over a quarter mile section and includes nine different identifying features for traffic at any given time. The initial step is to select the most pertinent of these features as a target for extraction and local storage during the experiment. The tools created for this approach, a two-level hierarchical group of operators, are used to extract features from the dataset to create a feature space; this is done to minimize the experimental set to a matrix of desirable attributes from the vehicles on the roadway. The application is to identify if this data can be readily parsed into four distinct traffic states; in this case, the state of a vehicle is defined by its velocity and acceleration at a selected timestamp. A three-dimensional plot is used, with color as the third dimension and seen from a top-down perspective, to initially identify vehicle states in a section of roadway over a selected section of time. This is followed by applying k-means clustering, in this case with k=4 to match the four distinct traffic states, to the feature space to examine its viability in determining the states of vehicles in a time section. The method’s accuracy is viewed through silhouette plots. Finally, a group of experiments run through a decision-tree architecture is compared to the kmeans clustering approach. Each decision-tree format uses sets of predefined values for velocity and acceleration to parse the data into the four states; modifications are made to acceleration and deceleration values to examine different results. The three-dimensional plots provide a visual example of congested traffic for use in performing visual comparisons of the clustering results. The silhouette plot results of the k-means experiments show inaccuracy for certain clusters; on the other hand, the decision-tree work shows promise for future work.

One experiment was adminstered to address three issues central to identifying the processes that underlie our ability to learn through observation. One objective of the study was to identify the movement features (relative or absolute) extracted by an observer when demonstration acts as the training protocol. A second objective was to investigate how the performance level of the model (trial-to-trial variability in strategy selection) providing the demonstrations influences movement feature extraction. Lastly, a goal was to test whether or not visual perspective of the model by the observer (first-person or third-person) interacts with the aforementioned variables. The goal of the task was to trace two circles templates with a 90 degree relative phase offset between the two hands. Video recordings of two models practicing over three days were used to make three videos for the study; an expert performance, discovery performance, and instruction performance video. The discovery video portrayed a decrease in relative phase error and a transition from high trial-to-trial variability in the strategy selection to use of a single strategy. The instruction video also portrayed a decrease in relative phase error, but with no strategy search throughout practice. The expert video showed no strategy search with trial-to-trial variability within 5% of the goal relative phase of 90 across every trial. Observers watched one of the three video recordings from either a first-person or third-person perspective. In a retention test, the expert observers showed the most consistant capability (learning) in performing the goal phase. The instruction observers also showed learning, but to a lesser degree than the expert observers. The discovery group observers showed the least amount of learning of relative phase. The absolute feature of movement amplitude was not extracted by any observer group, results consistent with postulations by Scully and Newell (1985). Observation from the 1P perspective proved optimal in the expert and instruction observation groups, but the 3P perspective allowed for greater learning of of the goal relative phase (90 degree) in the discovery observation group. Hand lead, a relative feature of motion, was extracted by most obsevers, except those who observed the discovery model from the 3P perspective. It's concluded that the trial-to-trial variabiliy in terms of strategy selection interacted with the process of mental rotation, which prevented the extraction of hand lead in those observers that viewed the discovery model.

No description available.