In this article, the cognitive vision module of an autonomous flying robot is studied. The problem of the scene understanding by the robot, which flies on the high altitude, is analyzed. In such conditions, the examined scene can be regarded as two-dimensional. It is assumed that the robot operates in the urban-type environment. The scene representation is stored in the neighborhood graph that collects data about the objects locations, shapes, and their spatial relations. The fragments of the scene are understood by the robot in the context of neighborhoods of the objects. It is shown that such information can be effectively used for recognition of the object, while many objects of similar shape exist in the scene. In the proposed recognition process, not only the information about the shape of the object is utilized but also the spatial relations with other objects in its close neighborhood are examined.
Abstract-In this paper the authors present the results of research to develop the visual system for autonomous flying agent. The core elements of the vision system which were designed and implemented in the earlier stage of the project are brought together. The second aim is to show capabilities of a simulation environment designed and developed by the authors in order to enable testing of the vision systems (dedicated for Unmanned Aerial Vehicles) in the artificial environment. The first section of the paper introduces the testing (simulation) environment for MavLink-protocol-based autonomous flying robots. Next, the core elements of a vision system, designed for Unmanned Aerial Vehicle (UAV), are discussed. This includes pre-processing and vectorization algorithms, object recognition methods and fast three-dimensional model construction. The third part introduces a set of algorithms for robot navigation, solely based on vision and altitude sensor and compass. The paper concludes with the description of the tests and presentation of results where designed simulator was applied to show mentioned vision system elements operating together to execute complex task.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.