Visual object localization for mobile robots

A visual object localization system for mobile robots is developed and evaluated within an integrated testing environment. The system, which consists of segmentation, tracking and looming modules, estimates the position of objects with respect to a robot moving freely in its environment.; The segmentation module segments the objects to be localized. It uses a novel algorithm developed for the automatic detection and segmentation of colored objects in real time. These objects are then tracked in time by the tracking module.; The looming module employs the visual looming method to localize objects using the displacement of the robot measured through its odometry and the change in size of the objects in the image due to the robot's displacement. The visual looming method assumes that the object is placed head-on and the displacement of the robot is towards the object. A novel extension of the method that improves the localization performance under arbitrary displacements of the robot is also developed and implemented in the looming module.; A focus of the present research is the development of an integrated environment for the evaluation of the visual object localization system. The integrated environment superimposes spatial information obtained from the robot's sensors upon an overhead view of the robot's environment. That view provides a common coordinate system for navigation and localization algorithms to be visually and quantitatively evaluated in real time.; Systematic experiments with the visual object localization system are carried out in the integrated environment. These experiments allow quantitative evaluation of the performance of the system under different circumstances. The noise sensitivity of the visual looming method is derived analytically, and results compared with experimental data.; The performance of the method is compared with that of two other visual localization methods, namely 1-D optical flow and running-fix triangulation. The performance of the visual looming method is also compared with that of sonar, a popular localization sensor for mobile robots, on a different mobile robot platform. The strengths and weaknesses of the various localization methods are then summarized.