The development of deep learning-based biometric models that can be deployed on devices with constrained memory and computational resources has proven to be a significant challenge. Previous approaches to this problem have not prioritized the reduction of feature map redundancy, but the introduction of Ghost modules represents a major innovation in this area. Ghost modules use a series of inexpensive linear transformations to extract additional feature maps from a set of intrinsic features, allowing for a more comprehensive representation of the underlying information. GhostNetV1 and GhostNetV2, both of which are based on Ghost modules, serve as the foundation for a group of lightweight face recognition models called GhostFaceNets. GhostNetV2 expands upon the original GhostNetV1 by adding an attention mechanism to capture long-range dependencies. Evaluation of GhostFaceNets using various benchmarks reveals that these models offer superior performance while requiring a computational complexity of approximately 60-275 MFLOPs. This is significantly lower than that of State-Of-The-Art (SOTA) big convolutional neural network (CNN) models, which can require hundreds of millions of FLOPs. GhostFaceNets trained with the ArcFace loss on the refined MS-Celeb-1M dataset demonstrate SOTA performance on all benchmarks. In comparison to previous SOTA mobile CNNs, GhostFaceNets greatly improve efficiency for face verification tasks. The GhostFaceNets entire code will be made available after publication.

Rotary-wing unmanned aerial vehicles (UAVs) have the ability to operate in confined spaces and to hover over point of interest, but they have limited flight time and endurance. Conventional contact-based charging system for UAVs has been used, but it requires high landing accuracy for proper docking. Instead of the conventional system, autonomous wireless battery charging system for UAVs in outdoor conditions is proposed in this paper. UAVs can be wirelessly charged using the proposed charging system, regardless of yaw angle between UAVs and wireless charging pad, which can further reduce their control complexity for autonomous landing. The increased overall mission time eventually relaxes the limitations on payload and flight time. In this paper, a cost effective automatic recharging solution for UAVs in outdoor environments is proposed using wireless power transfer (WPT). This research proposes a global positioning system (GPS) and vision-based closed-loop target detection and a tracking system for precise landing of quadcopters in outdoor environments. The system uses the onboard camera to detect the shape, color and position of the defined target in image frame. Based on the offset of the target from the center of the image frame, control commands are generated to track and maintain the center position. Commercially available AR.Drone. was used to demonstrate the proposed concept which is equppied with bottom camera and GPS. Experiments and analyses showed good performance, and about 75% average WPT efficiency was achieved in this research.

Vision-based localization systems, namely visual odometry (VO) and visual inertial odometry (VIO), have attracted great attention recently. They are regarded as critical modules for building fully autonomous systems. The simplicity of visual and inertial state estimators, along with their applicability in resource-constrained platforms motivated robotic community to research and develop novel approaches that maximize their robustness and reliability. In this paper, we surveyed state-of-the-art VO and VIO approaches. In addition, studies related to localization in visually degraded environments are also reviewed. The reviewed VO techniques and related studies have been analyzed in terms of key design aspects including appearance, feature, and learning based approaches. On the other hand, research studies related to VIO have been categorized based on the degree and type of fusion process into loosely-coupled, semi-tightly coupled, or tightly-coupled approaches and filtering or optimization-based paradigms. This paper provides an overview of the main components of visual localization, key design aspects highlighting the pros and cons of each approach, and compares the latest research works in this field. Finally, a detailed discussion of the challenges associated with the reviewed approaches and future research considerations are formulated.

Aerial manipulation has direct application prospects in environment, construction, forestry, agriculture, search, and rescue. It can be used to pick and place objects and hence can be used for transportation of goods. Aerial manipulation can be used to perform operations in environments inaccessible or unsafe for human workers. This paper is a survey of recent research in aerial manipulation. The aerial manipulation research has diverse aspects, which include the designing of aerial manipulation platforms, manipulators, grippers, the control of aerial platform and manipulators, the interaction of aerial manipulator with the environment, through forces and torque. In particular, the review paper presents the survey of the airborne platforms that can be used for aerial manipulation including the new aerial platforms with aerial manipulation capability. We also classified the aerial grippers and aerial manipulators based on their designs and characteristics. The recent contributions regarding the control of the aerial manipulator platform is also discussed. The environment interaction of aerial manipulators is also surveyed which includes, different strategies used for end-effectors interaction with the environment, application of force, application of torque and visual servoing. A recent and growing interest of researchers about the multi-UAV collaborative aerial manipulation was also noticed and hence different strategies for collaborative aerial manipulation are also surveyed, discussed and critically analyzed. Some key challenges regarding outdoor aerial manipulation and energy constraints in aerial manipulation are also discussed.