Visual Line Research Articles

Femtosecond Laser-Assisted Graft Preparation and Implantation of Corneal Allogeneic Intrastromal Ring Segments for Corneal Ectasia: 1-Year Results.

To evaluate the outcomes of patients treated with corneal allogeneic intrastromal ring segments cut with femtosecond laser (Femto-CAIRS) without concomitant corneal crosslinking. Patients with keratoconus treated with Femto-CAIRS at the American University of Beirut Medical Center were included (May 2022-January 2023). A proprietary software program was developed on the femtosecond laser to cut allogeneic segments. Visual, refractive, tomographic, aberrometric, and epithelial data by anterior segment optical coherence tomography were measured at baseline and 1, 3, 6, and 12 months postoperatively. 20 eyes of 15 patients were included and followed up for 12 months. The manifest refraction spherical equivalent and cylinder improved from -6.79 ± 4.9 diopter (D) and -4.25 ± 1.8 D to -1.88 ± 2.9 D (P < 0.001) and -2.64 ± 1.4 D (P = 0.01) 12 months postoperatively, respectively. 75% of eyes gained 3 or more corrected distance visual acuity lines, most of which (65%) gained 4 lines or more 12 months postoperatively. The maximum keratometry and vertical coma decreased by 5.2 D (P < 0.001) and 1 μm (P = 0.001), respectively, 3 months postoperatively and remained stable until 12 months. The largest anterior stromal elevation over the central 5-mm diameter decreased from 36.0 ± 18.2 μm preoperatively to 19.9 ± 9.25 μm at 1 week postoperatively (P < 0.001) and remained relatively stable. Epithelial thickness over the cone increased relative to baseline starting 1 month postoperatively and becoming stable after 6 months while the mean epithelial thickness central to the ring peaked at 1 month after which it decreased to reach a plateau at 6 months. The Femto-CAIRS procedure improves visual and tomographic parameters and allows repeatable and safe results with the possibility of customization for individualized management.

Tightly coupled stereo vision-inertial odometry based on point and line feature

Abstract To improve the accuracy and robustness of visual simultaneous localization and mapping (SLAM) in low-texture environments, this paper proposes a robust and fast stereo vision inertial SLAM pose estimation method that combines point and line features with an inertial measurement unit (IMU). The method tightly couples visual point and line features with IMU constraints, forming a least-squares problem through the error of each constraint term for nonlinear optimization. To address the issues of over-segmentation and time consumption in traditional line segment detection (LSD) algorithms, an improved LSD algorithm is adopted to accelerate line feature extraction. This approach merges nearby line segments based on spatial geometric relationships and filters out invalid segments, improving the time efficiency of the algorithm. Finally, experiments conducted in low-texture environments demonstrate that our algorithm achieves high localization accuracy and robustness.

Towards scalable wide area UAS monitoring of forest disturbance using hydrogen powered airships

ABSTRACT Uncrewed Aircraft Systems (UAS) have been recognized as powerful tools for providing monitoring data supporting mitigating and adapting to various abiotic and biotic forest disturbances. However, their suitability and commercial viability have been significantly limited by the necessity for operation primarily within visual line of sight (VLOS). Efforts are currently ongoing to establish frameworks for beyond visual line of sight (BVLOS) UAS operation. This study represents the first successful application of hydrogen-powered BVLOS airship UAS technology for monitoring forest disturbances. The test area covered 1.3 km2 of conserved forest area in North Karelia, Finland, seriously affected by bark beetles and other disturbance agents. Monitoring flights were conducted in spring, summer, and autumn in BVLOS setting operated from a command centre situated approximately 75 km distance from the test area. Deep learning models were applied to study the tree health in the disturbance area, with the special focus on examining the scalability of machine learning models. Multispectral imagery outperformed RGB imagery in classifying trees as healthy, infested, and dead spruce trees, as well as non-spruce tree classes. With multispectral imagery, the F1-scores were 0.88–0.94, 0.80, 0.92–0.96, and 0.75–0.82 for healthy, infested, and dead spruce trees, and non-spruces, respectively. Transfer learning and fine-tuning, which adapted previously trained models to new study areas, improved results compared to training models from scratch and thus contributed to the development of scalable UAS remote sensing methods in mitigating forest disturbances on a larger scale.

The level of skills involved in an observation-based gait analysis.

This study aimed to determine the visual assessment skills during an observation-based gait analysis. Participants (N=40) included 20 physiotherapists (PTs) with>10 years of clinical experience (physiotherapists) and 20 physiotherapy students. Both groups watched a video of the gait of a subject with Guillain-Barré syndrome before and after being provided with information regarding other movements. Further, visual lines were measured using an EMR-8 eye mark recorder, and the results were compared between both groups. The average gaze duration was longer for students than for PTs (F1,79=53.3; p<0.01), whereas PTs gazed more often than the students (F1,79=87.6; p< 0.01). Furthermore, the PTs moved their eyes vertically more often than the students (F1,151=9.1; P< 0.01). We found that being able to discriminate the relative physical relationship of body locations by frequent and rapid vertical gazes could be an indication of the level of skills as an index to express the visual assessment skill in an observation-based gait analysis.

UAV Visual and Thermographic Power Line Detection Using Deep Learning.

Inspecting and maintaining power lines is essential for ensuring the safety, reliability, and efficiency of electrical infrastructure. This process involves regular assessment to identify hazards such as damaged wires, corrosion, or vegetation encroachment, followed by timely maintenance to prevent accidents and power outages. By conducting routine inspections and maintenance, utilities can comply with regulations, enhance operational efficiency, and extend the lifespan of power lines and equipment. Unmanned Aerial Vehicles (UAVs) can play a relevant role in this process by increasing efficiency through rapid coverage of large areas and access to difficult-to-reach locations, enhanced safety by minimizing risks to personnel in hazardous environments, and cost-effectiveness compared to traditional methods. UAVs equipped with sensors such as visual and thermographic cameras enable the accurate collection of high-resolution data, facilitating early detection of defects and other potential issues. To ensure the safety of the autonomous inspection process, UAVs must be capable of performing onboard processing, particularly for detection of power lines and obstacles. In this paper, we address the development of a deep learning approach with YOLOv8 for power line detection based on visual and thermographic images. The developed solution was validated with a UAV during a power line inspection mission, obtaining mAP@0.5 results of over 90.5% on visible images and over 96.9% on thermographic images.

Image Segmentation-Based Oilseed Rape Row Detection for Infield Navigation of Agri-Robot

The segmentation and extraction of oilseed rape crop rows are crucial steps in visual navigation line extraction. Agricultural autonomous navigation robots face challenges in path recognition in field environments due to factors such as complex crop backgrounds and varying light intensities, resulting in poor segmentation and slow detection of navigation lines in oilseed rape crops. Therefore, this paper proposes VC-UNet, a lightweight semantic segmentation model that enhances the U-Net model. Specifically, VGG16 replaces the original backbone feature extraction network of U-Net, Convolutional Block Attention Module (CBAM) are integrated at the upsampling stage to enhance focus on segmentation targets. Furthermore, channel pruning of network convolution layers is employed to optimize and accelerate the model. The crop row trapezoidal ROI regions are delineated using end-to-end vertical projection methods with serialized region thresholds. Then, the centerline of oilseed rape crop rows is fitted using the least squares method. Experimental results demonstrate an average accuracy of 94.11% for the model and an image processing speed of 24.47 fps/s. After transfer learning for soybean and maize crop rows, the average accuracy reaches 91.57%, indicating strong model robustness. The average yaw angle deviation of navigation line extraction is 3.76°, with a pixel average offset of 6.13 pixels. Single image transmission time is 0.009 s, ensuring real-time detection of navigation lines. This study provides upper-level technical support for the deployment of agricultural robots in field trials.

Integrating Null Controllability and Model-Based Safety Assessment for Enhanced Reliability in Drone Design

The increasing use of drones for safety-critical applications, particularly beyond visual lines of sight and over densely populated areas, necessitates safer and more reliable designs. To address this need, this paper introduces a novel methodology integrating Null Controllability with the Model-Based Safety Assessment (MBSA) framework AltaRica 3.0 to optimize propulsor configurations and system architectures. The main advancement of this method lies in the automation of reliability modeling and the integration of controllability assessment, eliminating restrictions on the types of propulsor configurations and system architectures that can be evaluated and significantly reducing the effort required for each design iteration. Through a hexarotor drone case study, the proposed method enabled a high number of design iterations, efficiently exploring various aspects of the design problem simultaneously, such as configuration, system architecture, and controllability hypothesis, which is not possible with state-of-the-art techniques. This approach demonstrated significant reliability improvements by implementing and optimizing redundancies, reducing the probability of loss of control by up to 99%. The case study also highlighted the increasing difficulty of enhancing reliability with each iteration and confirmed that it is unnecessary to consider more than two simultaneous failures for design optimization. A comparison of reliability figures with previous studies highlights the crucial role of system architecture in effectively enhancing drone design reliability. This work advances the field by providing an effective multidisciplinary modeling framework for drone design, enhancing reliability in safety-critical applications.

Development and evaluation of the model for acute kidney injury in patients with cardiac arrest after successful resuscitation

BackgroundThis study aims to construct a clinical prediction model and create a visual line chart depicting the risk of acute kidney injury (AKI) following resuscitation in cardiac arrest (CA) patients. Additionally, the study aims to validate the clinical predictive accuracy of the developed model.MethodsData were retrieved from the Dryad database, and publicly shared data were downloaded. This retrospective cohort study included 347 successfully resuscitated patients post-cardiac arrest from the Dryad database. Demographic and clinical data of patients in the database, along with their renal function during hospitalization, were included. Through data analysis, the study aimed to explore the relevant influencing factors of acute kidney injury (AKI) in patients after cardiopulmonary resuscitation. The study constructed a line chart prediction model using multivariate logistic regression analysis with post-resuscitation shock status (Post-resuscitation shock refers to the condition where, following successful cardiopulmonary resuscitation after cardiac arrest, some patients develop cardiogenic shock.), C reactive protein (CRP), Lactate dehydrogenase (LDH), and Alkaline phosphatase (ALP) identified as predictive factors. The predictive efficiency of the fitted model was evaluated by the area under the curve (AUC) of the receiver operating characteristic (ROC) curve.ResultsMultivariate logistic regression analysis showed that post-resuscitation shock status, CRP, LDH, and PAL were the influencing factors of AKI after resuscitation in CA patients. The calibration curve test indicated that the prediction model was well-calibrated, and the results of the Decision Curve Analysis (DCA) demonstrated the clinical utility of the model constructed in this study.ConclusionPost-resuscitation shock status, CRP, LDH, and ALPare the influencing factors for AKI after resuscitation in CA patients. The clinical prediction model constructed based on the above indicators has good clinical discriminability and practicality.

An Intelligent Bait Delivery Control Method for Flight Vehicle Evasion Based on Reinforcement Learning

During aerial combat, when an aircraft is facing an infrared air-to-air missile strike, infrared baiting technology is an important means of penetration, and the strategy of effective delivery of infrared bait is critical. To address this issue, this study proposes an improved deep deterministic policy gradient (DDPG) algorithm-based intelligent bait-dropping control method. Firstly, by modeling the relative motion between aircraft, bait, and incoming missiles, the Markov decision process of aircraft-bait-missile infrared effect was constructed with visual distance and line of sight angle as states. Then, the DDPG algorithm was improved by means of pre-training and classification sampling. Significantly, the infrared bait-dropping decision network was trained through interaction with the environment and iterative learning, which led to the development of the bait-dropping strategy. Finally, the corresponding environment was transferred to the Nvidia Jetson TX2 embedded platform for comparative testing. The simulation results showed that the convergence speed of this method was 46.3% faster than the traditional DDPG algorithm. More importantly, it was able to generate an effective bait-throwing strategy, enabling the aircraft to successfully evade the attack of the incoming missile. The strategy instruction generation time is only about 2.5 ms, giving it the ability to make online decisions.

The effects of drone transportation on blood component quality: A prospective randomised controlled laboratory study.

The use of uncrewed aerial vehicles (drones) has increased over the last decade. However, their application in healthcare has not been fully examined, in part, due to regulations preventing flight beyond the visual line of sight. This prospective randomised controlled laboratory study aimed to determine whether the invitro quality of packed red blood cell components is maintained when transported by drone, beyond visual line of sight. Ten identical pairs of packed red blood cell units were randomly allocated to transport by drone or by ground vehicle (1:1, allocation concealment) 68 km between two hospitals in Northumbria, UK. Markers of blood component quality were compared at 8, 14, 28 and 35 days following blood unit manufacture. There was no statistical difference in haemolysis, potassium concentration, total haemoglobin, glucose and lactate, haematocrit and mean cell volume, between the two groups, up to the date of unit expiry. The temperature of the packed red blood cell units did not deviate outside the recommended 2-10°C for transportation, regardless of the allocated group. Blood component transport was faster by drone, but did not reach statistical significance. This study demonstrates the feasibility and safety of flying blood components by drone between hospitals in the United Kingdom.

Optimized Radio Frequency Footprint Identification Based on UAV Telemetry Radios.

With the widespread use of unmanned aerial vehicles (UAVs), the detection and identification of UAVs is a vital security issue for the safety of airspace and ground facilities in the no-fly zone. Telemetry radios are important wireless communication devices for UAVs, especially in UAVs beyond the visual line of sight (BVLOS) operating mode. This work focuses on the UAV identification approach using transient signals from UAV telemetry radios instead of the signals from UAV controllers that the former research work depended on. In our novel UAV Radio Frequency (RF) identification system framework based on telemetry radio signals, the EC-α algorithm is optimized to detect the starting point of the UAV transient signal and the detection accuracy at different signal-to-noise ratios (SNR) is evaluated. In the training stage, the Convolutional Neural Network (CNN) model is trained to extract features from raw I/Q data of the transient signals with different waveforms. Its architecture and hyperparameters are analyzed and optimized. In the identification stage, the extracted transient signals are clustered through the Self-Organizing Map (SOM) algorithm and the Clustering Signals Joint Identification (CSJI) algorithm is proposed to improve the accuracy of RF fingerprint identification. To evaluate the performance of our proposed approach, we design a testbed, including two UAVs as the flight platform, a Universal Software Radio Peripheral (USRP) as the receiver, and 20 telemetry radios with the same model as targets for identification. Indoor test results show that the optimized identification approach achieves an average accuracy of 92.3% at 30 dB. In comparison, the identification accuracy of SVM and KNN is 69.7% and 74.5%, respectively, at the same SNR condition. Extensive experiments are conducted outdoors to demonstrate the feasibility of this approach.

Factual Analysis of Factors Influencing Consumer Cognitive Thinking and Automobile Designing using Fuzzy-AHP

Consumer cognitive and emotional responses significantly impact consumers’ affinity for both tangible and intangible products. Recent research underscores the heightened influence of consumer emotional and cognitive perceptions in product design, surpassing reliance on designers’ instincts or experiences. This case study delved into the analysis of automobiles, using cars as focal product. While visual elements like curve lines, grill design, and color predominantly shape consumer perspectives, non-visual factors such as reliability, quality, and ergonomics play a pivotal role in purchasing decisions. Our investigation, involving six industrial experts and over 130 participants, explored non-visual factors affecting consumers’ cognitive perceptions during car purchasing. Through a two-phase experimental approach, we developed a framework to understand the disparities in ranking assigned by consumers and designers to these non-visual factors, revealing perception gaps. Utilizing the rank value methodology, the Pareto principle, and Fuzzy-AHP analysis, we identified mileage/fuel efficiency, safety features, and reliability as dominant factors influencing consumer perceptions. Designers, on the other hand, prioritize safety features, reliability, and quality/warranty. Our findings emphasize the strategic significance for automobile companies to enhance their success rates by prioritizing these key factors, ultimately elevating brand value, profile, and overall financial prosperity.

Risks of Multipath Signal Propagation From GPS Satellites in Mountainous Areas During the Use of UAV as Supplementary Device for HEMS or SAR

It cannot be disputed that using unmanned aerial vehicles as a supplementary search and rescue component is still increasingly being discussed. Several approaches to solving the problem can be stated - buying specially designed unmanned aerial vehicles for HEMS (Helicopter Emergency Medical Service)/SAR (Search and Rescue) or modifying conventional UAVs to supplementary devices for HEMS or SAR. Regardless of which alternative the end-user prefers, the navigation method and determining the current position in space stays unchanged - primarily, the navigation signals from GNSS (Global Navigation Satellite Systems) satellites are used. However, this method can be loaded by the error caused by environmental conditions in which the rescue UAV moves, such as mountains, valleys, specific flora, trees, and much more. All of that causes attenuation, loss or multipath propagation of the signal. In the case of the remote-controlled flight of the UAV (by a pilot-operator or operator itself from the ground station) and when the rule of direct visibility (Visual Line of Sight - VLOS) is used for an unmanned aerial vehicle, this may not be a considerable problem. However, during an automated or fully autonomous flight and the flight of an unmanned aerial vehicle beyond direct visibility (Beyond Visual Line of Sight - BVLOS), knowledge about the exact position of the UAV in space is crucial. The presented article is dedicated to verify the claim about the significance of the problem of multipath signal propagation from the GPS (Global Positioning System) satellite in a selected mountain area, The Little Cold Valley in High Tatras, Slovak Republic.

A Fusion Approach for UAV Onboard Flight Trajectory Management and Decision Making Based on the Combination of Enhanced A* Algorithm and Quadratic Programming

The rapid advancement of unmanned aerial vehicle (UAV) technologies has led to an increasing demand for UAV operations in low-altitude, high-density, and complex airspace such as mountains or urban areas. In order to handle complex scenarios and ensure flight safety for UAVs with different flight missions beyond visual line of sight in such environments, a fusion framework of onboard autonomous flight trajectory management and decision-making system using global strategical path planning and local tactical trajectory optimization combination is proposed in this paper. The global strategical path planning is implemented by an enhanced A* algorithm under the multi-constraint of UAV positioning uncertainty and obstacle density to improve the safety and cost-effectiveness. The local tactical trajectory optimization is realized using quadratic programming to ensure smoothness, kinematic feasibility, and obstacle avoidance of the planned trajectory in dynamic environments. Receding-horizon control is used to ensure the flight path and trajectory planning efficiently and seamlessly. To assess the performance of the system, a terrain database and a navigation system are employed for environment and navigation performance simulation. The experimental results confirm that the fusion approach can realize better safety and cost-effectiveness through path planning with kino-dynamic feasible trajectory optimization.

Cost-effectiveness of erosion mitigation to meet water clarity targets in the Manawatū-Whanganui Region of New Zealand

Soil erosion is a significant environmental issue worldwide. It affects water quality, biodiversity, and land productivity. New Zealand government agencies and regional councils work to mitigate soil erosion through policies, management programmes, and funding for soil conservation projects. Information about cost-effectiveness is crucial for planning, targeting, and implementing erosion mitigation to achieve improvements in sediment-related water quality. While there is a good understanding of the costs of erosion mitigation measures, there is a dearth of literature on their cost-effectiveness in reducing sediment loads and improving water quality at the catchment level. In this study, we estimate the cost-effectiveness of erosion mitigation measures in meeting visual water clarity targets. The analysis utilizes the spatially explicit SedNetNZ erosion process and sediment budget modelling in the Manawatū-Whanganui Region and region-specific mitigation costs. The erosion mitigation measures considered in the analysis include afforestation, bush retirement, riparian retirement, space-planted trees, and gully tree planting. We modelled two scenarios with on-farm erosion mitigation implemented across the region from 2021 to 2100, resulting in a 48% and 60% reduction of total sediment load. We estimate the marginal costs to achieve the visual national bottom line for water clarity, as assessed by the length of waterways that meet the clarity targets. We also estimate the marginal costs of improving average water clarity, which can be linked with non-market valuation studies when conducting a cost-benefit analysis. We find that gully tree planting and space-planted trees are the most cost-effective mitigation measures and that riparian retirement is the least cost-effective. Moreover, cost-effectiveness is highly dependent on current land use and the biophysical features of the landscape. Our estimates can be used in cost-benefit analysis to plan and prioritize soil erosion mitigation at the catchment and regional levels.

Detecting Line Segments in Motion-Blurred Images With Events.

Making line segment detectors more reliable under motion blurs is one of the most important challenges for practical applications, such as visual SLAM and 3D line mapping. Existing line segment detection methods face severe performance degradation for accurately detecting and locating line segments when motion blur occurs. While event data shows strong complementary characteristics to images for minimal blur and edge awareness at high-temporal resolution, potentially beneficial for reliable line segment recognition. To robustly detect line segments over motion blurs, we propose to leverage the complementary information of images and events. Specifically, we first design a general frame-event feature fusion network to extract and fuse the detailed image textures and low-latency event edges, which consists of a channel-attention-based shallow fusion module and a self-attention-based dual hourglass module. We then utilize the state-of-the-art wireframe parsing networks to detect line segments on the fused feature map. Moreover, due to the lack of line segment detection datasets with pairwise motion-blurred images and events, we contribute two datasets, i.e., synthetic FE-Wireframe and realistic FE-Blurframe, for network training and evaluation. Extensive analyses on the component configurations demonstrate the design effectiveness of our fusion network. When compared to the state-of-the-arts, the proposed approach achieves the highest detection accuracy while maintaining comparable real-time performance. In addition to being robust to motion blur, our method also exhibits superior performance for line detection under high dynamic range scenes.

Network Performance and Technological Feasibility of Unmanned Aerial Vehicles for Network Extension

The operational range of conventional and license-free radio-controlled drones is limited due to line-of-sight restrictions (LoS). There exists a definitive method for operating a drone. Consequently, in order to fly the drone beyond the visual line of sight (BVLoS), it is necessary to replace the drone's original wireless communications equipment with a device that requires a licence and is connected to a cellular network. Long-Term Evolution (LTE), a terrestrial communication technique, enables a drone to establish a real-time connection with a ground station. This connection serves the goals of command and control (C&amp;C) as well as payload delivery. Nevertheless, it is important to note that the electromagnetic environment undergoes changes as altitude increases, which can potentially complicate the process of interfacing with drones over terrestrial cellular networks. The objective of this article is to develop a prototype control system for low-altitude microdrones using LTE technology. Additionally, it seeks to assess the feasibility and effectiveness of cellular connectivity for drones operating at various altitudes. This evaluation will be conducted by examining factors like as latency, handover, and signal strength. At a certain altitude, the received signal experiences a decrease in power level by 20 dBm and a degradation in signal quality by 10 dB. The data throughput of the downlink had a fall of 70%, while the latency exhibited an increase of 94 ms. Despite meeting the basic criteria for drone cellular connection, the existing LTE network necessitates enhancements in order to expand aerial coverage, mitigate interference, and minimise network latency.

Implementing and testing a U-space system: lessons learnt

Within the framework of the European Union’s Horizon 2020 research and innovation program, one of the main goals of the Labyrinth project, which ended on May 2023, was to develop and test the Conflict Management services of a U-space-based Unmanned Traffic Management (UTM) system. The U-space ConOps provides a high-level description of the architecture, requirements and functionalities of these systems, but the implementer has a certain degree of freedom in aspects like the techniques used or some policies and procedures. The current document describes some of those implementation decisions. The prototype included, at least in a basic version, part of the services defined by the ConOps, namely e-identification, Tracking, Geo-awareness, Drone Aeronautical Information Management, Geo-fence Provision, Operation Plan Preparation/Optimization, Operation Plan Processing, Strategic and Tactical Conflict Resolution, Emergency Management, Monitoring, Traffic Information and Legal Recording. Besides, a web app interface was developed for the operator/pilot. The system was tested in simulations and real visual line of sight (VLOS) and beyond VLOS (BVLOS) flights, with both vertical take-off and landing (VTOL) and fixed-wing platforms, while assisting final users interested in incorporating drones to support their daily tasks. The three-year experience developing and testing the environment provided many lessons at different levels: functionalities, compatibility, procedures, information, usability, ground control station (GCS) integration and aircrew roles during the mission.

Analysis of the new market trends of UAVS for safe BVLOS operations

UAVs are currently conquering the skies as prominent tools for various data-intensive applications, in the economic, transport, military and civil sector. While initially found application in the military sector, technology progression allowed them to enter the recreational sector and are now gaining ground in the fringes of the commercial environment. In parallel, technical components and subsystems that are application-optimised focus on highly automated drones, benefit from expertise in other domains, especially when it comes to Electronic Components and Systems (ECS), such as the automotive one, to operate beyond the visual line of sight (BVLOS) with a rather high degree of autonomy. Such technological developments, as well as currents trends and societal needs have opened the way for an unparalleled expansion in the use of UAS for a great number of applications, where humans cannot reach or are unable to perform in a timely and efficient manner. This work aims to present a in-depth analysis of the current market trends that shape the existing landscape for the development of safe and reliable BVLOS operations.

On UAV Terrestrial Connectivity Enhancement through Smart Selective Antennas

Nowadays, Unmanned Aerial Vehicles (UAVs) are widely used in heterogeneous contexts and, thanks to a continuous technological evolution, are going to be used for several applications such as, for example, Beyond Visual Line of Sight (BVLOS) operations. Since in BVLOS flights the UAV and the ground control center may not have a direct visibility with each other, a robust communication system is needed to provide reliable connectivity. Although a cellular (4G/5G) network is the current best candidate to enable BVLOS applications, there are still some limitations to overcome, as 4G (LTE) and 5G (NR) cellular networks are natively designed for terrestrial use. In this paper, we first investigate current cellular communication limitations for UAV-based applications, in particular taking into account both results available in the literature, as well as experimental performance campaigns. Then, a viable solution for mitigating these drawbacks exploiting selective on-board antennas is proposed, whose performance is experimentally investigated with a preliminary prototypical architecture.