Características Visuales para el Reconocimiento Rápido de Objetos

FERNANDO ENRIQUE CORREA TOME

Please use this identifier to cite or link to this item: http://repositorio.ugto.mx/handle/20.500.12059/2223

Full metadata record

DC Field	Value	Language
dc.rights.license	http://creativecommons.org/licenses/by-nc-nd/4.0	es_MX
dc.contributor	RAUL ENRIQUE SANCHEZ YAÑEZ	es_MX
dc.creator	FERNANDO ENRIQUE CORREA TOME	es_MX
dc.date.accessioned	2020-08-05T18:05:23Z	-
dc.date.available	2020-08-05T18:05:23Z	-
dc.date.issued	2015-09	-
dc.identifier.uri	http://repositorio.ugto.mx/handle/20.500.12059/2223	-
dc.language.iso	eng	es_MX
dc.rights	info:eu-repo/semantics/openAccess	es_MX
dc.subject.classification	CIS- Doctorado en Ingeniería Eléctrica	es_MX
dc.title	Características Visuales para el Reconocimiento Rápido de Objetos	es_MX
dc.title.alternative	Visual Features for Fast Object Recognition	en
dc.type	info:eu-repo/semantics/doctoralThesis	es_MX
dc.creator.id	info:eu-repo/dai/mx/cvu/295697	es_MX
dc.subject.cti	info:eu-repo/classification/cti/7	es_MX
dc.subject.cti	info:eu-repo/classification/cti/33	es_MX
dc.subject.keywords	Object recognition	en
dc.subject.keywords	Precision-Recall Graphics	en
dc.subject.keywords	Algorithm Hopcroft-Karp	en
dc.subject.keywords	PHD (Partial Hausdorff Distance)	en
dc.subject.keywords	ISM Methodology (Integral Split and Merge)	en
dc.subject.keywords	Reconocimiento de objetos	es_MX
dc.subject.keywords	Gráficos Precision-Recall	es_MX
dc.subject.keywords	Algoritmo Hopcroft-Karp	es_MX
dc.subject.keywords	PHD (Distancia Parcial de Hausdorff)	es_MX
dc.subject.keywords	Metodología ISM (División y Fusión Integral)	es_MX
dc.contributor.id	info:eu-repo/dai/mx/cvu/30994	es_MX
dc.contributor.role	director	es_MX
dc.type.version	info:eu-repo/semantics/publishedVersion	es_MX
dc.publisher.university	Universidad de Guanajuato	es_MX
dc.description.abstractEnglish	In this thesis study, we explored the recognition of objects using visual features. From the different approaches we studied during the development of this work, three main research lines arise: a study of fast object recognition by shape, a fast segmentation methodology, and an online object recognition system that learns from example images. For the first study, a visual similarity metric of shapes, based on Precision-Recall graphs is presented, as an alternative to the widely used Hausdorff distance (HD). Such metric, called maximum cardinality similarity metric, is computed between a reference shape and a test template, each one represented by a set of edge points. We address this problem using a bipartite graph representation of the relationship between the sets. The matching problem is solved using the Hopcroft-Karp algorithm, taking advantage of its low computational complexity. We present a comparison between our results and those obtained from applying the partial Hausdorff distance (PHD) to the same test sets. Similar results were found using both approaches for standard template-matching applications. Nevertheless, the proposed methodology is more accurate at determining the completeness of partial shapes under noise conditions. Furthermore, the processing time required by our methodology is lower than that required to compute the PHD, for a large set of points. The second study presents a split-and-merge segmentation methodology, that uses integral images to improve the execution time. We call our methodology integral split and merge (ISM) segmentation. The integral images are used here to calculate statistics of the image regions in constant time. Those statistics are used to guide the splitting process by identifying the homogeneous regions in the image. We also propose a merge criterion that performs connected component analysis of the homogeneous regions. Moreover, the merging procedure is able to group regions of the image showing gradients. Furthermore, the number of regions resulting from the segmentation process is determined automatically. In a series of tests, we compare ISM against other state-of-the-art algorithms. The results from the tests show that our ISM methodology obtains image segmentations with a comparable quality, using a simple texture descriptor instead of a combination of color-texture descriptors. The proposed ISM methodology also has a piecewise linear computational complexity, resulting in an algorithm fast enough to be executed in real time. Finally, in the third study we present an object recognition system based on histograms of visual features, obtained from the images containing the objects to recognize. The system is divided in two stages, the feature extraction, and the classification stages. The first stage involves transforming the original image to the CIELuv color space, and then extracting four different visual features from this space. The features are: the color hue, the color saturation, the luminance, and a texture feature that is the standard deviation of the luminance, calculated for local neighborhoods. A histogram is then calculated from each feature and then merged together, in order to obtain a combined feature histogram called THSL histogram. Later, a collection of THSL histograms is built as our knowledge database, and is used for the classification of unseen images. This classification stage consists in the extraction of a THSL histogram from the image under test. This histogram is then compared against all the histograms in the knowledge database, and the label of the most similar histogram is retrieved as the classification result. We performed a series of tests, using images of small objects taken from different points of view. The images have no background, the objects are at the same distance from the camera, and the illumination is constant. The results obtained show a classification success rate around the 96% for individual objects, under the aforementioned conditions.	es_MX
Appears in Collections:	Doctorado en Ingeniería Eléctrica

Files in This Item:

File	Description	Size	Format
Correa Tome Fernando Enrique..pdf		1.81 MB	Adobe PDF	View/Open

Show simple item record