Computer Vision Techniques for Ambient Intelligence Applications

Ambient Intelligence (AmI) is a muldisciplinary area which refers to environments that are sensitive and responsive to the presence of people and objects. The rapid progress of technology and simultaneous reduction of hardware costs characterizing the recent years have enlarged the number of possible AmI applications, thus raising at the same time new research challenges. In particular, one important requirement in AmI is providing a proactive support to people in their everyday working and free-time activities. To this aim, Computer Vision represents a core research track since only through suitable vision devices and techniques it is possible to detect elements of interest and understand the occurring events. The goal of this thesis is presenting and demonstrating efficacy of novel machine vision research contributes for different AmI scenarios: object keypoints analysis for Augmented Reality purpose, segmentation of natural images for plant species recognition and heterogeneous people identification in unconstrained environments.