Computational Techniques for Spectral Image Analysis

Multispectral images are becoming more common in the field of remote sensing, computer vision, and industrial applications. Due to the high accuracy of the multispectral information, it can be used as an important quality factor in the inspection of industrial products. Recently, the development on multispectral imaging systems and the computational analysis on the multispectral images have been the focus of a growing interest. In this thesis, three areas of multispectral image analysis are considered. First, a method for analyzing multispectral textured images was developed. The method is based on a spectral cooccurrence matrix, which contains information of the joint distribution of spectral classes in a spectral domain. Next, a procedure for estimating the illumination spectrum of the color images was developed. Proposed method can be used, for example, in color constancy, color correction, and in the content based search from color image databases. Finally, color filters for the optical pattern recognition were designed, and a prototype of a spectral vision system was constructed. The spectral vision system can be used to acquire a low dimensional component image set for the two dimensional spectral image reconstruction. The data obtained by the spectral vision system is small and therefore convenient for storing and transmitting a spectral image.

Efficient image mosaicing for multi-robot visual underwater mapping

Robotic platforms have advanced greatly in terms of their remote sensing capabilities, including obtaining optical information using cameras. Alongside these advances, visual mapping has become a very active research area, which facilitates the mapping of areas inaccessible to humans. This requires the efficient processing of data to increase the final mosaic quality and computational efficiency. In this paper, we propose an efficient image mosaicing algorithm for large area visual mapping in underwater environments using multiple underwater robots. Our method identifies overlapping image pairs in the trajectories carried out by the different robots during the topology estimation process, being this a cornerstone for efficiently mapping large areas of the seafloor. We present comparative results based on challenging real underwater datasets, which simulated multi-robot mapping

Anàlisi d'imatges estereoscòpiques utilitzant mètodes de baixa càrrega computacional

Mitjançant imatges estereoscòpiques es poden detectar la posició respecte de la càmera dels objectes que apareixen en una escena. A partir de les diferències entre les imatges captades pels dos objectius es pot determinar la profunditat dels objectes. Existeixen diversitat de tècniques de visió artificial que permeten calcular la localització dels objectes, habitualment amb l’objectiu de reconstruir l’escena en 3D. Aquestes tècniques necessiten una gran càrrega computacional, ja que utilitzen mètodes de comparació bidimensionals, i per tant, no es poden utilitzar per aplicacions en temps real. En aquest treball proposem un nou mètode d’anàlisi de les imatges estereoscòpiques que ens permeti obtenir la profunditat dels objectes d’una escena amb uns resultats acceptables. Aquest nou mètode es basa en transformar la informació bidimensional de la imatge en una informació unidimensional per tal de poder fer la comparació de les imatges amb un baix cost computacional, i dels resultats de la comparació extreure’n la profunditat dels objectes dins l’escena. Això ha de permetre, per exemple, que aquest mètode es pugui implementar en un dispositiu autònom i li permeti realitzar operacions de guiatge a través d’espais interiors i exteriors.

Reconeixement de gestos de la mà amb el sensor Kinect

El reconeixement dels gestos de la mà (HGR, Hand Gesture Recognition) és actualment un camp important de recerca degut a la varietat de situacions en les quals és necessari comunicar-se mitjançant signes, com pot ser la comunicació entre persones que utilitzen la llengua de signes i les que no. En aquest projecte es presenta un mètode de reconeixement de gestos de la mà a temps real utilitzant el sensor Kinect per Microsoft Xbox, implementat en un entorn Linux (Ubuntu) amb llenguatge de programació Python i utilitzant la llibreria de visió artifical OpenCV per a processar les dades sobre un ordinador portàtil convencional. Gràcies a la capacitat del sensor Kinect de capturar dades de profunditat d’una escena es poden determinar les posicions i trajectòries dels objectes en 3 dimensions, el que implica poder realitzar una anàlisi complerta a temps real d’una imatge o d’una seqüencia d’imatges. El procediment de reconeixement que es planteja es basa en la segmentació de la imatge per poder treballar únicament amb la mà, en la detecció dels contorns, per després obtenir l’envolupant convexa i els defectes convexos, que finalment han de servir per determinar el nombre de dits i concloure en la interpretació del gest; el resultat final és la transcripció del seu significat en una finestra que serveix d’interfície amb l’interlocutor. L’aplicació permet reconèixer els números del 0 al 5, ja que s’analitza únicament una mà, alguns gestos populars i algunes de les lletres de l’alfabet dactilològic de la llengua de signes catalana. El projecte és doncs, la porta d’entrada al camp del reconeixement de gestos i la base d’un futur sistema de reconeixement de la llengua de signes capaç de transcriure tant els signes dinàmics com l’alfabet dactilològic.

Optimització de la tècnica del miralls màgics a partir de mesures de teoria de la informació

La visualització científica estudia i defineix algorismes i estructures de dades que permeten fer comprensibles conjunts de dades a través d’imatges. En el cas de les aplicacions mèdiques les dades que cal interpretar provenen de diferents dispositius de captació i es representen en un model de vòxels. La utilitat d’aquest model de vòxels depèn de poder-lo veure des del punt de vista ideal, és a dir el que aporti més informació. D’altra banda, existeix la tècnica dels Miralls Màgics que permet veure el model de vòxels des de diferents punts de vista alhora i mostrant diferents valors de propietat a cada mirall. En aquest projecte implementarem un algorisme que permetrà determinar el punt de vista ideal per visualitzar un model de vòxels així com també els punts de vista ideals per als miralls per tal d’aconseguir el màxim d’informació possible del model de vòxels. Aquest algorisme es basa en la teoria de la informació per saber quina és la millor visualització. L’algorisme també permetrà determinar l’assignació de colors òptima per al model de vòxels

Comparison of Local Feature Detectors and Descriptors for Visual Object Categorization

Local features are used in many computer vision tasks including visual object categorization, content-based image retrieval and object recognition to mention a few. Local features are points, blobs or regions in images that are extracted using a local feature detector. To make use of extracted local features the localized interest points are described using a local feature descriptor. A descriptor histogram vector is a compact representation of an image and can be used for searching and matching images in databases. In this thesis the performance of local feature detectors and descriptors is evaluated for object class detection task. Features are extracted from image samples belonging to several object classes. Matching features are then searched using random image pairs of a same class. The goal of this thesis is to find out what are the best detector and descriptor methods for such task in terms of detector repeatability and descriptor matching rate.

Bag-of-Features Approach to Unsupervised Visual Object Categorisation

The large and growing number of digital images is making manual image search laborious. Only a fraction of the images contain metadata that can be used to search for a particular type of image. Thus, the main research question of this thesis is whether it is possible to learn visual object categories directly from images. Computers process images as long lists of pixels that do not have a clear connection to high-level semantics which could be used in the image search. There are various methods introduced in the literature to extract low-level image features and also approaches to connect these low-level features with high-level semantics. One of these approaches is called Bag-of-Features which is studied in the thesis. In the Bag-of-Features approach, the images are described using a visual codebook. The codebook is built from the descriptions of the image patches using clustering. The images are described by matching descriptions of image patches with the visual codebook and computing the number of matches for each code. In this thesis, unsupervised visual object categorisation using the Bag-of-Features approach is studied. The goal is to find groups of similar images, e.g., images that contain an object from the same category. The standard Bag-of-Features approach is improved by using spatial information and visual saliency. It was found that the performance of the visual object categorisation can be improved by using spatial information of local features to verify the matches. However, this process is computationally heavy, and thus, the number of images must be limited in the spatial matching, for example, by using the Bag-of-Features method as in this study. Different approaches for saliency detection are studied and a new method based on the Hessian-Affine local feature detector is proposed. The new method achieves comparable results with current state-of-the-art. The visual object categorisation performance was improved by using foreground segmentation based on saliency information, especially when the background could be considered as clutter.

Hand tracking in high-speed camera videos

Visual object tracking has been one of the most popular research topics in the field of computer vision recently. Specifically, hand tracking has attracted significant attention since it would enable many useful practical applications. However, hand tracking is still a very challenging problem which cannot be considered solved. The fact that almost every aspect of hand appearance can change is the fundamental reason for this difficulty. This thesis focused on 2D-based hand tracking in high-speed camera videos. During the project, a toolbox for this purpose was collected which contains nine different tracking methods. In the experiments, these methods were tested and compared against each other with both high-speed videos recorded during the project and publicly available normal speed videos. The results revealed that tracking accuracies varied considerably depending on the video and the method. Therefore, no single method was clearly the best in all videos, but three methods, CT, HT, and TLD, performed better than the others overall. Moreover, the results provide insights about the suitability of each method to different types and situations of hand tracking.

Probabilistic position and orientation estimation of rigid objects using an RGB-D sensor

The recent emergence of low-cost RGB-D sensors has brought new opportunities for robotics by providing affordable devices that can provide synchronized images with both color and depth information. In this thesis, recent work on pose estimation utilizing RGBD sensors is reviewed. Also, a pose recognition system for rigid objects using RGB-D data is implemented. The implementation uses half-edge primitives extracted from the RGB-D images for pose estimation. The system is based on the probabilistic object representation framework by Detry et al., which utilizes Nonparametric Belief Propagation for pose inference. Experiments are performed on household objects to evaluate the performance and robustness of the system.

Local Features in Image and Video Processing - Object Class Matching and Video Shot Detection

The usage of digital content, such as video clips and images, has increased dramatically during the last decade. Local image features have been applied increasingly in various image and video retrieval applications. This thesis evaluates local features and applies them to image and video processing tasks. The results of the study show that 1) the performance of different local feature detector and descriptor methods vary significantly in object class matching, 2) local features can be applied in image alignment with superior results against the state-of-the-art, 3) the local feature based shot boundary detection method produces promising results, and 4) the local feature based hierarchical video summarization method shows promising new new research direction. In conclusion, this thesis presents the local features as a powerful tool in many applications and the imminent future work should concentrate on improving the quality of the local features.

Post-processing and analysis of tracked hand trajectories

In this thesis, the suitability of different trackers for finger tracking in high-speed videos was studied. Tracked finger trajectories from the videos were post-processed and analysed using various filtering and smoothing methods. Position derivatives of the trajectories, speed and acceleration were extracted for the purposes of hand motion analysis. Overall, two methods, Kernelized Correlation Filters and Spatio-Temporal Context Learning tracking, performed better than the others in the tests. Both achieved high accuracy for the selected high-speed videos and also allowed real-time processing, being able to process over 500 frames per second. In addition, the results showed that different filtering methods can be applied to produce more appropriate velocity and acceleration curves calculated from the tracking data. Local Regression filtering and Unscented Kalman Smoother gave the best results in the tests. Furthermore, the results show that tracking and filtering methods are suitable for high-speed hand-tracking and trajectory-data post-processing.

Automatic image-based identification of Saimaa ringed seals

The Saimaa ringed seal is one of the most endangered seals in the world. It is a symbol of Lake Saimaa and a lot of effort have been applied to save it. Traditional methods of seal monitoring include capturing the animals and installing sensors on their bodies. These invasive methods for identifying can be painful and affect the behavior of the animals. Automatic identification of seals using computer vision provides a more humane method for the monitoring. This Master's thesis focuses on automatic image-based identification of the Saimaa ringed seals. This consists of detection and segmentation of a seal in an image, analysis of its ring patterns, and identification of the detected seal based on the features of the ring patterns. The proposed algorithm is evaluated with a dataset of 131 individual seals. Based on the experiments with 363 images, 81\% of the images were successfully segmented automatically. Furthermore, a new approach for interactive identification of Saimaa ringed seals is proposed. The results of this research are a starting point for future research in the topic of seal photo-identification.

Generative Part-Based Gabor Object Detector

Object detection is a fundamental task of computer vision that is utilized as a core part in a number of industrial and scientific applications, for example, in robotics, where objects need to be correctly detected and localized prior to being grasped and manipulated. Existing object detectors vary in (i) the amount of supervision they need for training, (ii) the type of a learning method adopted (generative or discriminative) and (iii) the amount of spatial information used in the object model (model-free, using no spatial information in the object model, or model-based, with the explicit spatial model of an object). Although some existing methods report good performance in the detection of certain objects, the results tend to be application specific and no universal method has been found that clearly outperforms all others in all areas. This work proposes a novel generative part-based object detector. The generative learning procedure of the developed method allows learning from positive examples only. The detector is based on finding semantically meaningful parts of the object (i.e. a part detector) that can provide additional information to object location, for example, pose. The object class model, i.e. the appearance of the object parts and their spatial variance, constellation, is explicitly modelled in a fully probabilistic manner. The appearance is based on bio-inspired complex-valued Gabor features that are transformed to part probabilities by an unsupervised Gaussian Mixture Model (GMM). The proposed novel randomized GMM enables learning from only a few training examples. The probabilistic spatial model of the part configurations is constructed with a mixture of 2D Gaussians. The appearance of the parts of the object is learned in an object canonical space that removes geometric variations from the part appearance model. Robustness to pose variations is achieved by object pose quantization, which is more efficient than previously used scale and orientation shifts in the Gabor feature space. Performance of the resulting generative object detector is characterized by high recall with low precision, i.e. the generative detector produces large number of false positive detections. Thus a discriminative classifier is used to prune false positive candidate detections produced by the generative detector improving its precision while keeping high recall. Using only a small number of positive examples, the developed object detector performs comparably to state-of-the-art discriminative methods.

ReAlM - a system to manipulate relations

Given a heterogeneous relation algebra R, it is well known that the algebra of matrices with coefficient from R is relation algebra with relational sums that is not necessarily finite. When a relational product exists or the point axiom is given, we can represent the relation algebra by concrete binary relations between sets, which means the algebra may be seen as an algebra of Boolean matrices. However, it is not possible to represent every relation algebra. It is well known that the smallest relation algebra that is not representable has only 16 elements. Such an algebra can not be put in a Boolean matrix form.[15] In [15, 16] it was shown that every relation algebra R with relational sums and sub-objects is equivalent to an algebra of matrices over a suitable basis. This basis is given by the integral objects of R, and is, compared to R, much smaller. Aim of my thesis is to develop a system called ReAlM - Relation Algebra Manipulator - that is capable of visualizing computations in arbitrary relation algebras using the matrix approach.

Extending relAPS to first order logic

RelAPS is an interactive system assisting in proving relation-algebraic theorems. The aim of the system is to provide an environment where a user can perform a relation-algebraic proof similar to doing it using pencil and paper. The previous version of RelAPS accepts only Horn-formulas. To extend the system to first order logic, we have defined and implemented a new language based on theory of allegories as well as a new calculus. The language has two different kinds of terms; object terms and relational terms, where object terms are built from object constant symbols and object variables, and relational terms from typed relational constant symbols, typed relational variables, typed operation symbols and the regular operations available in any allegory. The calculus is a mixture of natural deduction and the sequent calculus. It is formulated in a sequent style but with exactly one formula on the right-hand side. We have shown soundness and completeness of this new logic which verifies that the underlying proof system of RelAPS is working correctly.