Abstract
Cameras are one of the most relevant sensors in autonomous robots. However, two of their challenges are to extract useful information from captured images, and to manage the small field of view of regular cameras. This paper proposes implementing a dynamic visual memory to store the information gathered from a moving camera on board a robot, followed by an attention system to choose where to look with this mobile camera, and a visual localization algorithm that incorporates this visual memory. The visual memory is a collection of relevant task-oriented objects and 3D segments, and its scope is wider than the current camera field of view. The attention module takes into account the need to reobserve objects in the visual memory and the need to explore new areas. The visual memory is useful also in localization tasks, as it provides more information about robot surroundings than the current instantaneous image. This visual system is intended as underlying technology for service robot applications in real people's homes. Several experiments have been carried out, both with simulated and real Pioneer and Nao robots, to validate the system and each of its components in office scenarios.
Highlights
Computer vision research is growing rapidly, both in robotics and in many other applications, from surveillance systems for security to the automatic acquisition of 3D models for Virtual Reality displays.The number of commercial applications is increasing, such as traffic monitoring, parking entrance control, augmented reality video games, and face recognition
Regular cameras typically have 60 degrees of scope. This would be good enough for visual control but a broader scope may improve robot responses in tasks like navigation, where the presence of obstacles in the robot’s surroundings should be taken into account even if they lie outside the current field of view
If the uncertainty on a 3D segment falls below a given threshold, it is deleted from visual memory
Summary
Computer vision research is growing rapidly, both in robotics and in many other applications, from surveillance systems for security to the automatic acquisition of 3D models for Virtual Reality displays. The control algorithm should be robust enough to face the lack of time persistence of relevant stimuli in images This poses a challenge when the objects lie beyond the current field of view of the camera. The visual representation of interesting objects around the robot beyond the current field of view may improve the quality of the robot’s behavior as it handles more information when making decisions This is where the problem of selecting where to look every time, known as gaze control, or overt attention [2,3] arises. A visual localization algorithm has been developed which uses the current image or the contents of the memory to continuously estimate the robot position It provides a robust localization estimation and has been designed to handle symmetries in the environment.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
