Characterization and modelling of complex motion patterns

Movement analysis is the principle of any interaction with the world and the survival of living beings completely depends on the effciency of such analysis. Visual systems have remarkably developed eficient mechanisms that analyze motion at different levels, allowing to recognize objects in dynamical and cluttered environments. In artificial vision, there exist a wide spectrum of applications for which the study of complex movements is crucial to recover salient information. Yet each domain may be different in terms of scenarios, complexity and relationships, a common denominator is that all of them require a dynamic understanding that captures the relevant information. Overall, current strategies are highly dependent on the appearance characterization and usually they are restricted to controlled scenarios. This thesis proposes a computational framework that is inspired in known motion perception mechanisms and structured as a set of modules. Each module is in due turn composed of a set of computational strategies that provide qualitative and quantitative descriptions of the dynamic associated to a particular movement. Diverse applications were herein considered and an extensive validation was performed for each of them. Each of the proposed strategies has shown to be reliable at capturing the dynamic patterns of different tasks, identifying, recognizing, tracking and even segmenting objects in sequences of video.

• Autores: Martínez Carrillo, Fabio
• Tipo de recurso: Doctoral thesis
• Fecha de publicación: 2014
• Institución: Universidad Nacional de Colombia
• Repositorio: Universidad Nacional de Colombia
• Idioma: spa
• OAI Identifier: oai:repositorio.unal.edu.co:unal/52030
• Acceso en línea: https://repositorio.unal.edu.co/handle/unal/52030; http://bdigital.unal.edu.co/46280/
• Palabra clave: 0 Generalidades / Computer science, information and general works; 62 Ingeniería y operaciones afines / Engineering; Motion analysis; Dense Optical Flow; Background substraction; Action recognition; Tracking, gait analysis; Polyp detection; Hummingbird ight patterns; Cardiac MRI analysis; Video-surveillance; Análisis de movimiento; Flujo optico denso; Extracción de fondo; Reconocimiento de acciones; Seguimiento; Análisis de marcha; Detección de polipos; Patrones de vuelo del colibrí; Análisis de sequencias cardiacas MRI; Video-vigilancia
• Rights: openAccess
• License: Atribución-NoComercial 4.0 Internacional