• support vector machines
• robotica per l'assistenza personale
• DirectSVM
• SVM
• visione artificiale

In robotic systems, the use of vision, together with other perception systems, is motivated by the need to increase the versatility and the number of performable applications. If proximity, feel and force perception are significant to increase the performance of executions in traditional manufactory applications, vision is acknowledged as the most powerful among the perceptual abilities of a service robot.
The purpose of this work is to develop a system providing an assistive robot1 with visual capabilities for object localization and recognition based on a learning technique that increases the flexibility and robustness of the system and its generalization capacity. This work was almost entirely developed within the research group developing the Care-O-bot [30] system at the Fraunhofer IPA2, in order to explore the potentialities of a recent learning technique (the Support Vector Machines) to solve the problem of object detection and recognition for such a system.