The problem addressed in this thesis is the detection, localisation and tracking of human workers from mobile industrial machinery using a customised vision system developed at Örebro University. Coined the RefleX Vision System, its hardware configuration and computer vision algorithms were specifically designed for real-world industrial scenarios where workers are required to wear protective high-visibility garments with retro-reflective markers. The demand for robust industry-purpose human sensing methods originates from the fact that many industrial environments represent work spaces that are shared between humans and mobile machinery. Typical examples of such environments include construction sites, surface and underground mines, storage yards and warehouses. Here, accidents involving mobile equipment and human workers frequently result in serious injuries and fatalities. Robust sensor-based detection of humans in the surrounding of mobile equipment is therefore an active research topic and represents a crucial requirement for safe vehicle operation and accident prevention in increasingly automated production sites. Addressing the described safety issue, this thesis presents a collection of papers which introduce, analyse and evaluate a novel vision-based method for detecting humans equipped with protective high-visibility garments in the neighbourhood of manned or unmanned industrial vehicles. The thesis provides a comprehensive discussion of the numerous aspects regarding the design of the hardware and the computer vision algorithms that constitute the vision system. An active nearinfrared camera setup that is customised for the robust perception of retroreflective markers builds the basis for the sensing method. Using its specific input, a set of computer vision and machine learning algorithms then perform extraction, analysis, classification and localisation of the observed reflective patterns, and eventually detection and tracking of workers with protective garments. Multiple real-world challenges, which existing methods frequently struggle to cope with, are discussed throughout the thesis, including varying ambient lighting conditions and human body pose variation. The presented work has been carried out with a strong focus on industrial applicability, and therefore includes an extensive experimental evaluation in a number of different real-world indoor and outdoor work environments.

Universidade de Uberaba / In the past years, motion capture has been widely used in many application areas. In movies and games, motion capture is frequently employed to animate virtual characters. In sports, motion capture and analyses focus on optimizing movements of athletes and injury prevention. More applications areas include medicine, military and engineering. Motion capture can be accomplished by several technologies. However, optical marker-based systems are considered the gold standard of the motion capture field. They can offer high precision levels and flexibility to support most applications, but they are also the most expensive systems due to high costs of hardware and software. Although low-cost optical systems have been proposed in the last decade, these systems cannot provide enough precision, flexibility, automatism and/or real-time capability for a number of applications. In this context, the present research aims to develop a complete and high precision approach to track and reconstruct a cloud of independent markers, in real time, using multiple infrared specialized cameras. The proposed method includes a set of relatively simple algorithms which are part of a three-stage procedure. These stages work on the tracking and matching of the image points and spatial reconstruction of the marker trajectories. In order to evaluate the method, a prototype software has been implemented and experiments were performed using a pack of eight infrared cameras. The NaturalPoint s OptiTrack system, which includes the Arena software, was used as a reference system. In the experiments, the proposed technique was able to successfully track and reconstruct a set of 38 reflective markers in real time. When compared to the commercial software, the method performed better for automatically maker tracking. In addition, the reconstructed trajectories produced by the prototype software were far less contaminated by noise than the trajectories generated by the Arena software. The proposed method should encourage the development of new high performance systems at a more affordable price. / Nos últimos anos, os sistemas de captura de movimento vêm sendo aplicados em diversas áreas do conhecimento. A animação de personagens virtuais na indústria cinematográfica e a avaliação dos movimentos corporais em áreas da saúde e nos esportes são apenas algumas de suas diversas aplicações. Dos diferentes tipos de sistemas de captura de movimento existentes, os sistemas ópticos baseados em marcadores são reconhecidos como os mais avançados e os que oferecem níveis de precisão e flexibilidade suficientes para suportar o maior número de aplicações. Todavia, esses sistemas estão também entre os mais caros, devido ao alto preço dos elementos de hardware e software. As propostas de baixo custo existentes possuem limitações com relação à precisão, flexibilidade, automatismo e/ou velocidade de processamento, que as tornam inviáveis para uma considerável parcela de aplicações. Neste contexto, esta pesquisa propõe um método computacional completo, considerando as principais deficiências dos trabalhos existentes, para rastrear e reconstruir as trajetórias de uma nuvem de marcadores reflexivos independentes, em tempo real, utilizando múltiplas câmeras especializadas de infravermelho. A técnica proposta envolve um conjunto de algoritmos relativamente simples, os quais integram três etapas principais. Estas etapas realizam o rastreamento e o casamento dos pontos nas imagens e a reconstrução tridimensional das trajetórias dos marcadores. O método foi implementado em software e testado utilizando uma coleção de oito câmeras de infravermelho. Como referência, utilizou-se o software Arena e o sistema OptiTrack comercializado pela empresa NaturalPoint. Nos experimentos conduzidos, a técnica foi capaz de rastrear automaticamente um conjunto de 38 marcadores em tempo real. Comparada ao software comercial Arena, a técnica revelou uma melhor capacidade de rastreamento automático dos marcadores e ainda reconstruiu trajetórias com menor incidência de ruídos. O método proposto deve incentivar pesquisadores e empresas a desenvolverem sistemas de captura de movimento de alto desempenho e com menor custo. / Doutor em Ciências