Survey on path and view planning for UAVs

Background In recent decades,unmanned aerial vehicles(UAVs)have developed rapidly and been widely applied in many domains,including photography,reconstruction,monitoring,and search and rescue.In such applications,one key issue is path and view planning,which tells UAVs exactly where to fly and how to search.Methods With specific consideration for three popular UAV applications(scene reconstruction,environment exploration,and aerial cinematography),we present a survey that should assist researchers in positioning and evaluating their works in the context of existing solutions.Results/Conclusions It should also help newcomers and practitioners in related fields quickly gain an overview of the vast literature.In addition to the current research status,we analyze and elaborate on advantages,disadvantages,and potential explorative trends for each application domain.

A Survey of Blue-Noise Sampling and Its Applications

In this paper, we survey recent approaches to blue-noise sampling and discuss their beneficial applications. We discuss the sampling algorithms that use points as sampling primitives and classify the sampling algorithms based on various aspects, e.g., the sampling domain and the type of algorithm. We demonstrate several well-known applications that can be improved by recent blue-noise sampling techniques, as well as some new applications such as dynamic sampling and blue-noise remeshing.

Fitting boxes to Manhattan scenes using linear integer programming

We propose an approach for automatic generation of building models by assembling a set of boxes using a Manhattan-world assumption.The method first aligns the point cloud with a per-building local coordinate system,and then fits axis-aligned planes to the point cloud through an iterative regularization process.The refined planes partition the space of the data into a series of compact cubic cells(candidate boxes)spanning the entire 3D space of the input data.We then choose to approximate the target building by the assembly of a subset of these candidate boxes using a binary linear programming formulation.The objective function is designed to maximize the point cloud coverage and the compactness of the final model.Finally,all selected boxes are merged into a lightweight polygonal mesh model,which is suitable for interactive visualization of large scale urban scenes.Experimental results and a comparison with state-of-the-art methods demonstrate the effectiveness of the proposed framework.

Recurrent 3D attentional networks for end-to-end active object recognition

Active vision is inherently attention-driven:an agent actively selects views to attend in order to rapidly perform a vision task while improving its internal representation of the scene being observed.Inspired by the recent success of attention-based models in 2D vision tasks based on single RGB images, we address multi-view depth-based active object recognition using an attention mechanism, by use of an end-to-end recurrent 3D attentional network. The architecture takes advantage of a recurrent neural network to store and update an internal representation. Our model,trained with 3D shape datasets, is able to iteratively attend the best views targeting an object of interest for recognizing it. To realize 3D view selection, we derive a 3D spatial transformer network. It is dierentiable,allowing training with backpropagation, and so achieving much faster convergence than the reinforcement learning employed by most existing attention-based models. Experiments show that our method, with only depth input, achieves state-of-the-art next-best-view performance both in terms of time taken and recognition accuracy.