Classificação híbrida: pixel a pixel e baseada em objetos para o monitoramento da condição da superfície dos pavimentos rodoviários

Monitorar a condição de uso de toda a extensão das rodovias brasileiras é tarefa dispendiosa e demorada. Este trabalho trata de novas técnicas que permitem o levantamento da condição da superfície dos pavimentos rodoviários de forma ágil utilizando imagens hiperespectrais de sensor digital aeroembarcado. Nos últimos anos, um número crescente de imagens de alta resolução espacial tem surgido no mercado mundial com o aparecimento dos novos satélites e sensores aeroembarcados de sensoriamento remoto. Propõe-se uma metodologia para identificação dos pavimentos asfálticos e classificação das principais ocorrências dos defeitos na superfície do pavimento. A primeira etapa da metodologia é a identificação da superfície asfáltica na imagem, utilizando uma classificação híbrida baseada inicialmente em pixel e depois refinada por objetos. A segunda etapa da metodologia é a identificação e classificação das ocorrências dos principais defeitos nos pavimentos flexíveis que são observáveis nas imagens de alta resolução espacial. Esta última etapa faz uso intensivo das novas técnicas de classificação de imagens baseadas em objetos. O resultado final é a geração de índices da condição da superfície do pavimento a partir das imagens que possam ser comparados com os indicadores vigentes da condição da superfície do pavimento já normatizados pelos órgãos competentes no país.

Hierarchical multi-label classification using local neural networks

Hierarchical multi-label classification is a complex classification task where the classes involved in the problem are hierarchically structured and each example may simultaneously belong to more than one class in each hierarchical level. In this paper, we extend our previous works, where we investigated a new local-based classification method that incrementally trains a multi-layer perceptron for each level of the classification hierarchy. Predictions made by a neural network in a given level are used as inputs to the neural network responsible for the prediction in the next level. We compare the proposed method with one state-of-the-art decision-tree induction method and two decision-tree induction methods, using several hierarchical multi-label classification datasets. We perform a thorough experimental analysis, showing that our method obtains competitive results to a robust global method regarding both precision and recall evaluation measures.

Informationsgehalt von ERS-1/-2 SAR-Daten zur Erfassung der Arteninventare und des Zustandes landwirtschaftlich genutzter Böden und Vegetation

Auf einer drei Anbauperioden umfassenden Ground Truth Datenbasis wird der Informationsgehalt multitemporaler ERS-1/-2 Synthetic Aperture Radar (SAR) Daten zur Erfassung der Arteninventare und des Zustandes landwirtschaftlich genutzter Böden und Vegetation in Agrarregionen Bayerns evaluiert.Dazu wird ein für Radardaten angepaßtes, multitemporales, auf landwirtschaftlichen Schlägen beruhendes Klassifizierungsverfahren ausgearbeitet, das auf bildstatistischen Parametern der ERS-Zeitreihen beruht. Als überwachte Klassifizierungsverfahren wird vergleichend der Maximum-Likelihood-Klassifikator und ein Neuronales-Backpropagation-Netz eingesetzt. Die auf Radarbildkanälen beruhenden Gesamtgenauigkeiten variieren zwischen 75 und 85%. Darüber hinaus wird gezeigt, daß die interferometrische Kohärenz und die Kombination mit Bildkanälen optischer Sensoren (Landsat-TM, SPOT-PAN und IRS-1C-PAN) zur Verbesserung der Klassifizierung beitragen. Gleichermaßen können die Klassifizierungsergebnisse durch eine vorgeschaltete Grobsegmentierung des Untersuchungsgebietes in naturräumlich homogene Raumeinheiten verbessert werden. Über die Landnutzungsklassifizierung hinaus, werden weitere bio- und bodenphysikalische Parameter aus den SAR-Daten anhand von Regressionsmodellen abgeleitet. Im Mittelpunkt stehen die Paramter oberflächennahen Bodenfeuchte vegetationsfreier/-armer Flächen sowie die Biomasse landwirtschaftlicher Kulturen. Die Ergebnisse zeigen, daß mit ERS-1/-2 SAR-Daten eine Messung der Bodenfeuchte möglich ist, wenn Informationen zur Bodenrauhigkeit vorliegen. Hinsichtlich der biophysikalischen Parameter sind signifikante Zusammenhänge zwischen der Frisch- bzw. Trockenmasse des Vegetationsbestandes verschiedener Getreide und dem Radarsignal nachweisbar. Die Biomasse-Informationen können zur Korrektur von Wachstumsmodellen genutzt werden und dazu beitragen, die Genauigkeit von Ertragsschätzungen zu steigern.

Imaging and modeling of pore scale processes in porous media using X-ray computed tomography and lattice Boltzmann solver

In der Erdöl– und Gasindustrie sind bildgebende Verfahren und Simulationen auf der Porenskala im Begriff Routineanwendungen zu werden. Ihr weiteres Potential lässt sich im Umweltbereich anwenden, wie z.B. für den Transport und Verbleib von Schadstoffen im Untergrund, die Speicherung von Kohlendioxid und dem natürlichen Abbau von Schadstoffen in Böden. Mit der Röntgen-Computertomografie (XCT) steht ein zerstörungsfreies 3D bildgebendes Verfahren zur Verfügung, das auch häufig für die Untersuchung der internen Struktur geologischer Proben herangezogen wird. Das erste Ziel dieser Dissertation war die Implementierung einer Bildverarbeitungstechnik, die die Strahlenaufhärtung der Röntgen-Computertomografie beseitigt und den Segmentierungsprozess dessen Daten vereinfacht. Das zweite Ziel dieser Arbeit untersuchte die kombinierten Effekte von Porenraumcharakteristika, Porentortuosität, sowie die Strömungssimulation und Transportmodellierung in Porenräumen mit der Gitter-Boltzmann-Methode. In einer zylindrischen geologischen Probe war die Position jeder Phase auf Grundlage der Beobachtung durch das Vorhandensein der Strahlenaufhärtung in den rekonstruierten Bildern, das eine radiale Funktion vom Probenrand zum Zentrum darstellt, extrahierbar und die unterschiedlichen Phasen ließen sich automatisch segmentieren. Weiterhin wurden Strahlungsaufhärtungeffekte von beliebig geformten Objekten durch einen Oberflächenanpassungsalgorithmus korrigiert. Die Methode der „least square support vector machine” (LSSVM) ist durch einen modularen Aufbau charakterisiert und ist sehr gut für die Erkennung und Klassifizierung von Mustern geeignet. Aus diesem Grund wurde die Methode der LSSVM als pixelbasierte Klassifikationsmethode implementiert. Dieser Algorithmus ist in der Lage komplexe geologische Proben korrekt zu klassifizieren, benötigt für den Fall aber längere Rechenzeiten, so dass mehrdimensionale Trainingsdatensätze verwendet werden müssen. Die Dynamik von den unmischbaren Phasen Luft und Wasser wird durch eine Kombination von Porenmorphologie und Gitter Boltzmann Methode für Drainage und Imbibition Prozessen in 3D Datensätzen von Böden, die durch synchrotron-basierte XCT gewonnen wurden, untersucht. Obwohl die Porenmorphologie eine einfache Methode ist Kugeln in den verfügbaren Porenraum einzupassen, kann sie dennoch die komplexe kapillare Hysterese als eine Funktion der Wassersättigung erklären. Eine Hysterese ist für den Kapillardruck und die hydraulische Leitfähigkeit beobachtet worden, welche durch die hauptsächlich verbundenen Porennetzwerke und der verfügbaren Porenraumgrößenverteilung verursacht sind. Die hydraulische Konduktivität ist eine Funktion des Wassersättigungslevels und wird mit einer makroskopischen Berechnung empirischer Modelle verglichen. Die Daten stimmen vor allem für hohe Wassersättigungen gut überein. Um die Gegenwart von Krankheitserregern im Grundwasser und Abwässern vorhersagen zu können, wurde in einem Bodenaggregat der Einfluss von Korngröße, Porengeometrie und Fluidflussgeschwindigkeit z.B. mit dem Mikroorganismus Escherichia coli studiert. Die asymmetrischen und langschweifigen Durchbruchskurven, besonders bei höheren Wassersättigungen, wurden durch dispersiven Transport aufgrund des verbundenen Porennetzwerks und durch die Heterogenität des Strömungsfeldes verursacht. Es wurde beobachtet, dass die biokolloidale Verweilzeit eine Funktion des Druckgradienten als auch der Kolloidgröße ist. Unsere Modellierungsergebnisse stimmen sehr gut mit den bereits veröffentlichten Daten überein.

Classification of intervertebral disk degeneration with axial T2 mapping

OBJECTIVE: The aim of this study was to establish an MRI classification system for intervertebral disks using axial T2 mapping, with a special focus on evaluating early degenerative intervertebral disks. MATERIALS AND METHODS: Twenty-nine healthy volunteers (19 men, 10 women; age range, 20-44 years; mean age, 31.8 years) were studied, and axial T2 mapping was performed for the L3-L4, L4-L5, and L5-S1 intervertebral disks. Grading was performed using three classification systems for degenerative disks: our system using axial T2 mapping and two other conventional classification systems that focused on the signal intensity of the nucleus pulposus or the structural morphology in sagittal T2-weighted MR images. We analyzed the relationship between T2, which is known to correlate with change in composition of intervertebral disks, and degenerative grade determined using the three classification systems. RESULTS: With axial T2 mapping, differences in T2 between grades I and II were smaller and those between grades II and III, and between grades III and IV, were larger than those with the other grading systems. The ratio of intervertebral disks classified as grade I was higher with the conventional classification systems than that with axial T2 mapping. In contrast, the ratio of intervertebral disks classified as grade II or III was higher with axial T2 mapping than that with the conventional classification systems. CONCLUSION: Axial T2 mapping provides a more T2-based classification. The new system may be able to detect early degenerative changes before the conventional classification systems can.

HOG-Like gradient-based descriptor for visual vehicle detection

One of the main challenges for intelligent vehicles is the capability of detecting other vehicles in their environment, which constitute the main source of accidents. Specifically, many methods have been proposed in the literature for video-based vehicle detection. Most of them perform supervised classification using some appearance-related feature, in particular, symmetry has been extensively utilized. However, an in-depth analysis of the classification power of this feature is missing. As a first contribution of this paper, a thorough study of the classification performance of symmetry is presented within a Bayesian decision framework. This study reveals that the performance of symmetry-based classification is very limited. Therefore, as a second contribution, a new gradient-based descriptor is proposed for vehicle detection. This descriptor exploits the known rectangular structure of vehicle rears within a Histogram of Gradients (HOG)-based framework. Experiments show that the proposed descriptor outperforms largely symmetry as a feature for vehicle verification, achieving classification rates over 90%.

Region-dependent vehicle classification using PCA features

Video-based vehicle detection is the focus of increasing interest due to its potential towards collision avoidance. In particular, vehicle verification is especially challenging due to the enormous variability of vehicles in size, color, pose, etc. In this paper, a new approach based on supervised learning using Principal Component Analysis (PCA) is proposed that addresses the main limitations of existing methods. Namely, in contrast to classical approaches which train a single classifier regardless of the relative position of the candidate (thus ignoring valuable pose information), a region-dependent analysis is performed by considering four different areas. In addition, a study on the evolution of the classification performance according to the dimensionality of the principal subspace is carried out using PCA features within a SVM-based classification scheme. Indeed, the experiments performed on a publicly available database prove that PCA dimensionality requirements are region-dependent. Hence, in this work, the optimal configuration is adapted to each of them, rendering very good vehicle verification results.

A study of feature combination for vehicle detection based on image processing

Video analytics play a critical role in most recent traffic monitoring and driver assistance systems. In this context, the correct detection and classification of surrounding vehicles through image analysis has been the focus of extensive research in the last years. Most of the pieces of work reported for image-based vehicle verification make use of supervised classification approaches and resort to techniques, such as histograms of oriented gradients (HOG), principal component analysis (PCA), and Gabor filters, among others. Unfortunately, existing approaches are lacking in two respects: first, comparison between methods using a common body of work has not been addressed; second, no study of the combination potentiality of popular features for vehicle classification has been reported. In this study the performance of the different techniques is first reviewed and compared using a common public database. Then, the combination capabilities of these techniques are explored and a methodology is presented for the fusion of classifiers built upon them, taking into account also the vehicle pose. The study unveils the limitations of single-feature based classification and makes clear that fusion of classifiers is highly beneficial for vehicle verification.

Improving kNN multi-label classification in Prototype Selection scenarios using class proposals

Prototype Selection (PS) algorithms allow a faster Nearest Neighbor classification by keeping only the most profitable prototypes of the training set. In turn, these schemes typically lower the performance accuracy. In this work a new strategy for multi-label classifications tasks is proposed to solve this accuracy drop without the need of using all the training set. For that, given a new instance, the PS algorithm is used as a fast recommender system which retrieves the most likely classes. Then, the actual classification is performed only considering the prototypes from the initial training set belonging to the suggested classes. Results show that this strategy provides a large set of trade-off solutions which fills the gap between PS-based classification efficiency and conventional kNN accuracy. Furthermore, this scheme is not only able to, at best, reach the performance of conventional kNN with barely a third of distances computed, but it does also outperform the latter in noisy scenarios, proving to be a much more robust approach.

Query Enrichment for Image Collections by Reuse of Classification Rules

User queries over image collections, based on semantic similarity, can be processed in several ways. In this paper, we propose to reuse the rules produced by rule-based classifiers in their recognition models as query pattern definitions for searching image collections.

Habitat Map for Lizard Island reef, Australia derived from a photo-transect survey field data collected in December 2011 and September/October 2012

An object based image analysis approach (OBIA) was used to create a habitat map of the Lizard Reef. Briefly, georeferenced dive and snorkel photo-transect surveys were conducted at different locations surrounding Lizard Island, Australia. For the surveys, a snorkeler or diver swam over the bottom at a depth of 1-2m in the lagoon, One Tree Beach and Research Station areas, and 7m depth in Watson's Bay, while taking photos of the benthos at a set height using a standard digital camera and towing a surface float GPS which was logging its track every five seconds. The camera lens provided a 1.0 m x 1.0 m footprint, at 0.5 m height above the benthos. Horizontal distance between photos was estimated by fin kicks, and corresponded to a surface distance of approximately 2.0 - 4.0 m. Approximation of coordinates of each benthic photo was done based on the photo timestamp and GPS coordinate time stamp, using GPS Photo Link Software (www.geospatialexperts.com). Coordinates of each photo were interpolated by finding the gps coordinates that were logged at a set time before and after the photo was captured. Dominant benthic or substrate cover type was assigned to each photo by placing 24 points random over each image using the Coral Point Count excel program (Kohler and Gill, 2006). Each point was then assigned a dominant cover type using a benthic cover type classification scheme containing nine first-level categories - seagrass high (>=70%), seagrass moderate (40-70%), seagrass low (<= 30%), coral, reef matrix, algae, rubble, rock and sand. Benthic cover composition summaries of each photo were generated automatically in CPCe. The resulting benthic cover data for each photo was linked to GPS coordinates, saved as an ArcMap point shapefile, and projected to Universal Transverse Mercator WGS84 Zone 56 South. The OBIA class assignment followed a hierarchical assignment based on membership rules with levels for "reef", "geomorphic zone" and "benthic community" (above).

A GEOBIA methodology for fragmented agricultural landscapes

Very high resolution remotely sensed images are an important tool for monitoring fragmented agricultural landscapes, which allows farmers and policy makers to make better decisions regarding management practices. An object-based methodology is proposed for automatic generation of thematic maps of the available classes in the scene, which combines edge-based and superpixel processing for small agricultural parcels. The methodology employs superpixels instead of pixels as minimal processing units, and provides a link between them and meaningful objects (obtained by the edge-based method) in order to facilitate the analysis of parcels. Performance analysis on a scene dominated by agricultural small parcels indicates that the combination of both superpixel and edge-based methods achieves a classification accuracy slightly better than when those methods are performed separately and comparable to the accuracy of traditional object-based analysis, with automatic approach.

Apports de la texture multibande dans la classification orientée-objets d'images multisources (optique et radar).

Résumé : La texture dispose d’un bon potentiel discriminant qui complète celui des paramètres radiométriques dans le processus de classification d’image. L’indice Compact Texture Unit (CTU) multibande, récemment mis au point par Safia et He (2014), permet d’extraire la texture sur plusieurs bandes à la fois, donc de tirer parti d’un surcroît d’informations ignorées jusqu’ici dans les analyses texturales traditionnelles : l’interdépendance entre les bandes. Toutefois, ce nouvel outil n’a pas encore été testé sur des images multisources, usage qui peut se révéler d’un grand intérêt quand on considère par exemple toute la richesse texturale que le radar peut apporter en supplément à l’optique, par combinaison de données. Cette étude permet donc de compléter la validation initiée par Safia (2014) en appliquant le CTU sur un couple d’images optique-radar. L’analyse texturale de ce jeu de données a permis de générer une image en « texture couleur ». Ces bandes texturales créées sont à nouveau combinées avec les bandes initiales de l’optique, avant d’être intégrées dans un processus de classification de l’occupation du sol sous eCognition. Le même procédé de classification (mais sans CTU) est appliqué respectivement sur : la donnée Optique, puis le Radar, et enfin la combinaison Optique-Radar. Par ailleurs le CTU généré sur l’Optique uniquement (monosource) est comparé à celui dérivant du couple Optique-Radar (multisources). L’analyse du pouvoir séparateur de ces différentes bandes à partir d’histogrammes, ainsi que l’outil matrice de confusion, permet de confronter la performance de ces différents cas de figure et paramètres utilisés. Ces éléments de comparaison présentent le CTU, et notamment le CTU multisources, comme le critère le plus discriminant ; sa présence rajoute de la variabilité dans l’image permettant ainsi une segmentation plus nette, une classification à la fois plus détaillée et plus performante. En effet, la précision passe de 0.5 avec l’image Optique à 0.74 pour l’image CTU, alors que la confusion diminue en passant de 0.30 (dans l’Optique) à 0.02 (dans le CTU).

Umbrella pine above ground biomass estimation with very high spatial resolution satellite images

Above ground biomass is frequently estimated with forest inventory data and an extrapolation method for the per unit area evaluations. This procedure is labour demanding and costly. In this study above ground biomass functions, whose independent variable is crown horizontal projection, were developed. Multi-resolution segmentation method and object-oriented classification, based on very high spatial resolution satellite images, were used to obtain the area of tree crown horizontal projection for umbrella pine (Pinus pinea L.). A set of inventory plots were measured and with existing allometric functions for this species above ground biomass per tree and per plot were calculated. The two data sets were used to fit linear functions both for individual plot and their cumulative values. The results show a good performance of the models. Errors smaller than 10% are obtained for stand areas greater than 1.4 ha. These functions have the advantages of estimating above ground biomass for all the area under study or surveillance, not requiring forest inventory; allow monitoring in short time periods; and are easily implemented in a geographical information system environment.

Learning objects recommendation in an adaptive educational hypermedia system

The aim of this paper is presenting the modules of the Adaptive Educational Hypermedia System PCMAT, responsible for the recommendation of learning objects. PCMAT is an online collaborative learning platform with a constructivist approach, which assesses the user’s knowledge and presents contents and activities adapted to the characteristics and learning style of students of mathematics in basic schools. The recommendation module and search and retrieval module choose the most adequate learning object, based on the user's characteristics and performance, and in this way contribute to the system’s adaptability.