Les mines sous marines sont très utilisées dans les conflits. Pour contrer cette menace, les marines s'équipent de moyens de lutte anti-mine autonomes afin d'éviter l'intervention humaine. Une mission de guerre des mines se découpe en quatre étapes distinctes : la détection des objets, la classification et l'identification puis la neutralisation. Cette thèse propose des solutions algorithmiques pour l'étape d'identification par caméra vidéo. Le drone d'identification connaît la position approximative de l'objet à identifier. La première mission de ce drone est de re-détecter l'objet avant de le classifier et de l'identifier. Le milieu sous-marin perturbe les images acquises par la caméra (absorption, diffusion). Pour faciliter la détection et la reconnaissance (classification et identification), nous avons prétraité les images. Nous avons proposé deux méthodes de détection des objets. Tout d'abord nous modifions le spectre de l'image afin d'obtenir une image dans laquelle il est possible de détecter les contours des objets. Une seconde méthode a été développée à partir de la soustraction du fond, appris en début de séquence vidéo. Les résultats obtenus avec cette seconde méthode ont été comparés à une méthode existante. Lorsqu'il y a une détection, nous cherchons à reconnaître l'objet. Pour cela, nous utilisons la corrélation. Les images de référence ont été obtenues à partir d'images de synthèse 3D des mines. Pour les différentes méthodes utilisées, nous avons optimisés les résultats en utilisant les informations de navigation. En effet, selon les déplacements du drone, nous pouvons fixer des contraintes qui vont améliorer la détection et réduire le temps de calcul nécessaire à l'identification. / To avoid the underwater mine threat and to limit human interventions, navies use autonomous underwater vehicles. An underwater mine warfare mission is divided into four steps : object detection, classification, identification and neutralization. This PhD thesis proposes algorithmic solutions for the identification step done with a video camera. Thanks to the detection step, the identification vehicle knows approximately the object position. First, the vehicle has to detect and position this object exactly. Then it will be classified and identified. The underwater medium affects the images acquired with a video camera through absorption and scattering. The first step of our algorithm is to preprocess the images to help the detection and recognition (classification and identification) steps.We have proposed two detection methods. The first one consists in modifying image spectrum in order to obtain an image in which we will be able to detect edges of objects. The second method, based on region segmentation, has been developed from background subtraction methods. The background image has been learned at the beginning of the video when there is no object. The results of the latter have been compared to those obtained with a state-of-art method, on data acquired at sea. Once we have detected an object, we want to recognize it. For that, we use the correlation technique. The reference images have been obtained from 3D computer generated images of mines. This novel approach gives promising results. For each developed method, we have optimized the results through the use of navigational information. Indeed, depending on vehicle's motion, we can set constraints to improve the detection step and reduce processing time.

Submitted by Kamila Costa (kamilavasconceloscosta@gmail.com) on 2015-06-11T20:01:02Z No. of bitstreams: 1 Dissertação-Daily D de O Rodrigues.pdf: 2391223 bytes, checksum: 06b57d0d17da9e4844b2d8482ac25cb0 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-06-15T18:06:50Z (GMT) No. of bitstreams: 1 Dissertação-Daily D de O Rodrigues.pdf: 2391223 bytes, checksum: 06b57d0d17da9e4844b2d8482ac25cb0 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-06-15T18:08:28Z (GMT) No. of bitstreams: 1 Dissertação-Daily D de O Rodrigues.pdf: 2391223 bytes, checksum: 06b57d0d17da9e4844b2d8482ac25cb0 (MD5) / Made available in DSpace on 2015-06-15T18:08:28Z (GMT). No. of bitstreams: 1 Dissertação-Daily D de O Rodrigues.pdf: 2391223 bytes, checksum: 06b57d0d17da9e4844b2d8482ac25cb0 (MD5) Previous issue date: 2015-02-27 / FAPEAM - Fundação de Amparo à Pesquisa do Estado do Amazonas / The enhancement of underwater images in applications in the area of Amazonian rivers has been increasingly required and needs further study especially where the rivers have high turbidity and low light. There is increasingly demand for automatic enhancement methods to carry out monitoring of fauna and flora intensive rivers, as well as for the maintenance of pipelines and underwater cables. The enhancement methods specified, developed and validated for using in the rivers of the Amazonia are faced with the problem of imaging quality. The research related to underwater am environments of the Amazon has to dead with high turbidity of the water, caused mainly due to particles in suspension and interaction of light with the environment. The underwater images extraction with satisfiable visibility of the environments of Amazonian rivers has become extremely indispensable and relevant, given that there are natural treasures still unexplored into the depths of these rivers, as well as there is need to maintain the underwater part of the transportation system gas LPG (Liquefied Petroleum Gas) Coari-Manaus. Given this promising scenario, this study aims to improve these images by applying techniques of enhancement using nonlinear filters, which promote the minimization of the light interaction characteristics with the environment, loss of contrast and color in images extracted from turbid underwater environments. The method was experimentally validated with images acquired from simulations of underwater scenes and images acquired in outdoor underwater environments. The proposed method is compared to two other techniques of highlighting or enhancement of images. As in this study, these techniques also require a single image as input. The results return images with enhanced visual quality, considering a large set of experiments with simulation data and real outdoors scenes. / O realce de imagens subaquáticas em aplicações na região dos rios amazônicos é cada vez mais requisitado e carece de um estudo mais aprofundado especialmente nos casos em que os rios apresentam alto índice de turbidez e baixa luminosidade. Estes rios têm demandado cada vez mais métodos de realce automáticos que realizem o monitoramento de sua fauna e flora, bem como manutenção de dutos e cabos subaquáticos. Os métodos de realce especificados, desenvolvidos e validados para uso nos rios da região, se deparam com o problema da qualidade de captação de imagens. As pesquisas relacionadas aos ambientes subaquáticos da Amazônia são prejudicadas pelo alto nível de turbidez de suas águas, causadas principalmente devido às partículas em suspensão e à interação da luz com o meio. A extração de imagens subaquáticas de visibilidade adequada aos ambientes dos rios amazônicos em geral, tem se demonstrado imprescindível e relevante, haja vista que, existem tesouros naturais ainda inexplorados nas profundezas desses rios. Por outro lado, verifica-se a necessidade de manutenção da parte subaquática do sistema de transporte de gás GLP (Gás Liquefeito de Petróleo) Coari-Manaus. Diante deste cenário promissor, este trabalho objetiva a melhoria dessas imagens através da aplicação de técnicas de realce com uso de filtros não lineares, que promovam a minimização das características da interação da luz com o meio, perda de contraste e cores em imagens extraídas de ambientes subaquáticos turvos. O método proposto é comparado a duas outras técnicas de realce ou melhoria de imagens que, como neste trabalho, também requerem uma única imagem como entrada. Os resultados obtidos retornam imagens com melhor qualidade visual, considerando-se um grande conjunto de experimentos realizados com dados de simulação e cenas reais obtidas em ambientes externos.