English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In the project Autonomous City Explorer (ACE) a mobile robot should autonomously, efficiently and safely navigate in unstructured urban environments. From the biological aspect, the robot should not only plan its visual attention to acquire essential information about the unknown real world but also estimate the ego motion for the navigation based on vision with definitely fulfilled real-time capability. To achieve this, a multi-camera system is developed, which contains a multi-focal multi-camera platform, the camera head, for attentional gaze control and two high-speed cameras mounted towards the grounds for accurate visual odometry in extreme terrain. How to apply the human visual attention selection model on a mobile robot has become an intensively investigated research field. An active vision system should autonomously plan the robot's view direction not only based on a specific task but also for stimulus-based exploration of unknown real-world environment to collect more information. Moreover, psychological experiments also show that the familiarity of the current context also strongly influences the human attention selection behavior. To solve this context-based attention selection problem, we propose a view direction planning strategy based on the information theory. This strategy combines top-down attention selection in 3D space and bottom-up attention selection on the basis of a 2D saliency map. In both spaces the information content increases are defined. The optimal view direction is chosen which results in a maximum information gain after a camera view direction change. The main contribution is that a concerted information-based scalar is inserted to evaluate the information gains in the both sides. Moreover, the robot behavior, the choice of attention selection mechanism, can be adaptive to the current context. In addition, we implemented the compute-intensive bottom-up attention on Graphics Processing Units (GPUs) using the Compute Unified Device Architecture (CUDA), which provides an excellent speed-up of the system, due to the highly parallelizable structure. Using 4 NVIDIA GeForce 8800 (GTX) graphics cards for the input images at a resolution of 640 x 480, the computational cost is only 3.1ms with a frame rate of 313 fps. The saliency map generation on GPUs is approximately 8.5 times faster than the standard CPU implementations. O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m

Towards High-Speed Vision for Attention and Navigation of Autonomous City Explorer (ACE)