Nowadays, vision systems in portable computers are becoming an important tool for personal use. Vision systems for object localization are an active area of research. This dissertation proposes an algorithm to locate position and objects in a regular environment with the use of a simple webcam and a personal computer. To that end, we use two algorithms of marker tracking to reboot often a Visual Simultaneous Localisation and Mapping algorithm. This dissertation also presents an implementation and a set of tests that validate the proposed algorithm. 

Indoor localization poses is a challenge to computer vision research, since one may not make use of GPS-based devices. A classic approach commonly used in museums, research institutes, etc, is the use of fiducial marker to track the users position. However, this approach is intrusive into the ambient and not always possible. A possible solution would be natural marker detection, but algorithms for this, such as SURF, have not yet achieved real-time performance. A promising approach is a Visual Simultaneous Localization and Mapping (VSLAM) algorithm, which, starting from a known position, is capable of generating a map of the surrounding environment in portable systems. The problem of SLAM algorithms is theirs error accumulation that builds up during the movement. This work presents an algorithm to locate 3D positions in non-instrumented indoor environments using a web camera. We define a hybrid approach, using a pattern-recognition method to reinitialize whenever possible a VSLAM algorithm. An implementation of the proposed algorithm use well-known computer vision algorithms,such as SURF and Davison's SLAM. In addition, tests were made on datasets from walks inside a room. Results indicate that our approach is better than a fiducial marker tracking and pure SLAM tracking in our test environment.

This paper presents an algorithm for using projective invariant patterns in augmented reality applications. It is actually an adaptation of a previous algorithm for an optical tracking device, that works with infrared illumination and filtering. The present algorithm removes the necessity of working in a controlled environment, which would be inadequate for augmented reality applications. In order to compensate the excess of image noise caused by the absence of the infrared system, the proposed algorithm includes a fast binary decision tree in the process flow. We show that the algorithm achieves real time rates.

Fast corner detector algorithms are important for achieving real time in different computer vision applications. In this paper, we present new algorithm implementations for corner detection that makes use of the graphics processing units (GPU) of commodity hardware. The programmable capabilities of modern GPUs make it possible to speed up counterpart CPU algorithms. For corner detector algorithms, most steps are easily translated from CPU to GPU. However, the feature selection step imposes challenges for being mapped to the GPU parallel computational model. This paper presents a template for implementing corner detector algorithms that run entirely on the GPU, resulting in significant speed-ups. The proposed template is used to implement the KLT corner detector and the Harris corner detector, and numerical results are shown to demonstrate the algorithm efficiency.

This paper presents MOCT, a multi-object chromatic tracking technique for real-time natural video processing. Its main step is the MOCT localization algorithm, that performs local data evaluations in order to apply a multiple output parallel reduction operator to the image. The reduction operator is used to localize the positions of the object centroids, to compute the number of pixels occupied by an object and its bounding boxes, and to update object trajectories in image space. The operator is analyzed using three di erent computation layouts and tested over several reduction factors.

The tracking of objects for Virtual and Augmented Reality applications normally considers either the use of markers or markerless techniques, based on characteristics of the object to be tracked. The approach based on markers is more established, but requires the insertion of these "artificial" elements in the tracked object. Markerless techniques, on the other hand, are still in an experimental phase, offering several challenges to be overcome. This paper presents a hybrid system that tries to get the best of both approaches. Invariant properties of retroreflexive spherical markers patterns are used to detect the markers in the object. The inclusion of these markers in known polygonal areas of the tracked objects helps the detection of intrinsic characteristics of them, providing more robustness to the tracking process. A case study using a construction in ruins mock-up is presented.

This work presents a system for visualization of archaeological artifacts using Augmented Reality. The proposed system is portable and uses commodity hardware. Portability aims to take care of itinerant expositions where the system needs to be easily mounted and remounted. To validate our proposal we implemented a system using robust algorithms for camera calibration, feature detection, correlation, elimination of outliers and homography calculation. Results are presented to validate our proposal e to draw some conclusions.