The recent emergence of deep learning for characterizing complex patterns in remote sensing imagery reveals its high potential to address some classic challenges in this domain, e.g. scene classification. Typical deep learning models require extremely large datasets with rich contents to train a multi-layer structure in order to capture the essential features of scenes. Compared with the benchmark datasets used in popular deep learning frameworks, however, the volumes of available remote sensing datasets are particularly limited, which have restricted deep learning methods from achieving full performance gains. In order to address this fundamental problem, this paper introduces a methodology to not only enhance the volume and completeness of training data for any remote sensing datasets, but also exploit the enhanced datasets to train a deep convolutional neural network (CNN) that achieves state-of-the-art scene classification performance. Specifically, we propose to enhance any original dataset by applying three operations: flip, translation and rotation to generate augmented data, and use the augmented dataset to train and obtain a more descriptive deep model. The proposed methodology is validated in three recently released remote sensing datasets, and confirmed as an effective technique that significantly contributes to potentially revolutionary changes in remote sensing scene classification, empowered by deep learning.
Unmanned aerial vehicles (UAVs) play an invaluable role in information collection and data fusion. Because of their mobility and the complexity of deployed environments, constant position awareness and collision avoidance are essential. UAVs may encounter and/or cause danger if their Global Positioning System (GPS) signal is weak or unavailable. This paper tackles the problem of constant positioning and collision avoidance on UAVs in outdoor (wildness) search scenarios by using received signal strength (RSS) from the on-board communication module. Colored noise is found in the RSS, which invalidates the unbiased assumptions in Least Square (LS) algorithms which are widely used in RSS based position estimation. A colored noise model is thus proposed and applied in the extended Kalman filter for distance estimation. Furthermore, the constantly changing path loss factor during UAV flight can also affect the accuracy of estimation. In order to overcome this challenge, we present an adaptive algorithm to estimate the path loss factor. Given the position and velocity information, if a collision is detected we further employ an orthogonal rule to adapt the UAV predefined trajectory. Theoretical results prove that such an algorithm can provide effective modification to satisfy the required performance. Experiments have confirmed the advantages of the proposed algorithms.
Abstract-The application of small civilian unmanned aerial vehicles (UAVs) has attracted great interest for disaster sensing. However, the limited computational capability and low energy resource of UAVs present a significant challenge to real-time data processing, networking and policy making, which are of vital importance to many disaster related applications such as oilspill detection and flooding. In order to address the challenges imposed by the sheer volume of captured data, particularly video data, the intermittent and limited network resources, and the limited resources on UAVs, a new cloud-supported UAV application framework has been proposed and a prototype system of such framework has been implemented in this paper. The framework integrates video acquisition, data scheduling, data offloading and processing, and network state measurement to deliver an efficient and scalable system. The prototype of the framework comprises of a client-side set of components hosted on the UAV which selectively offloads the captured data to a cloudbased server. The server provides real-time data processing and information feedback services to the incident control centre and client device/operator. Results of the prototype system are presented to demonstrate the feasibility of such framework.
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.