Modelling and Control Design of a Non-Collaborative UAV Wireless Charging System.

This study proposes an analytical model of a WPT system with three orthogonal transmitter coils organised to produce a concentrated and controlled omnidirectional magnetic field suited for charging a moving, rotating load, providing maximal energy transfer without receiving end feedback. In order to create a realistic 3D WPT simulation system and a precise controller design, the mutual coupling values in terms of the receiver angular positions are modelled using the Ansys software. In using the established model of the 3DWPT system, an extremum seeking control (ESC) is used to maximize the power transfer utilizing the input power as an objective function assigned with specified parametric values defining the WPT model. The output power transmitted by the sending-end coils to a load of a moving UAV rotating in orbit is displayed. According to simulation results, when the receiver UAV speed is close to 2250 deg/s, the controller can accomplish a maximum power transfer of 2.6w in almost 1ms.

Method for Determining Treated Metal Surface Quality Using Computer Vision Technology.

Computer vision and image processing techniques have been extensively used in various fields and a wide range of applications, as well as recently in surface treatment to determine the quality of metal processing. Accordingly, digital image evaluation and processing are carried out to perform image segmentation, identification, and classification to ensure the quality of metal surfaces. In this work, a novel method is developed to effectively determine the quality of metal surface processing using computer vision techniques in real time, according to the average size of irregularities and caverns of captured metal surface images. The presented literature review focuses on classifying images into treated and untreated areas. The high computation burden to process a given image frame makes it unsuitable for real-time system applications. In addition, the considered current methods do not provide a quantitative assessment of the properties of the treated surfaces. The markup, processed, and untreated surfaces are explored based on the entropy criterion of information showing the randomness disorder of an already treated surface. However, the absence of an explicit indication of the magnitude of the irregularities carries a dependence on the lighting conditions, not allowing to explicitly specify such characteristics in the system. Moreover, due to the requirement of the mandatory use of specific area data, regarding the size of the cavities, the work is challenging in evaluating the average frequency of these cavities. Therefore, an algorithm is developed for finding the period of determining the quality of metal surface treatment, taking into account the porous matrix, and the complexities of calculating the surface tensor. Experimentally, the results of this work make it possible to effectively evaluate the quality of the treated surface, according to the criterion of the size of the resulting irregularities, with a frame processing time of 20 ms, closely meeting the real-time requirements.

A Deep-Learning Model for Real-Time Red Palm Weevil Detection and Localization.

Background and motivation: Over the last two decades, particularly in the Middle East, Red Palm Weevils (RPW, Rhynchophorus ferruginous) have proved to be the most destructive pest of palm trees across the globe. Problem: The RPW has caused considerable damage to various palm species. The early identification of the RPW is a challenging task for good date production since the identification will prevent palm trees from being affected by the RPW. This is one of the reasons why the use of advanced technology will help in the prevention of the spread of the RPW on palm trees. Many researchers have worked on finding an accurate technique for the identification, localization and classification of the RPW pest. This study aimed to develop a model that can use a deep-learning approach to identify and discriminate between the RPW and other insects living in palm tree habitats using a deep-learning technique. Researchers had not applied deep learning to the classification of red palm weevils previously. Methods: In this study, a region-based convolutional neural network (R-CNN) algorithm was used to detect the location of the RPW in an image by building bounding boxes around the image. A CNN algorithm was applied in order to extract the features to enclose with the bounding boxes-the selection target. In addition, these features were passed through the classification and regression layers to determine the presence of the RPW with a high degree of accuracy and to locate its coordinates. Results: As a result of the developed model, the RPW can be quickly detected with a high accuracy of 100% in infested palm trees at an early stage. In the Al-Qassim region, which has thousands of farms, the model sets the path for deploying an efficient, low-cost RPW detection and classification technology for palm trees.

An Efficient and Effective Deep Learning-Based Model for Real-Time Face Mask Detection.

Since December 2019, the COVID-19 pandemic has led to a dramatic loss of human lives and caused severe economic crises worldwide. COVID-19 virus transmission generally occurs through a small respiratory droplet ejected from the mouth or nose of an infected person to another person. To reduce and prevent the spread of COVID-19 transmission, the World Health Organization (WHO) advises the public to wear face masks as one of the most practical and effective prevention methods. Early face mask detection is very important to prevent the spread of COVID-19. For this purpose, we investigate several deep learning-based architectures such as VGG16, VGG19, InceptionV3, ResNet-101, ResNet-50, EfficientNet, MobileNetV1, and MobileNetV2. After these experiments, we propose an efficient and effective model for face mask detection with the potential to be deployable over edge devices. Our proposed model is based on MobileNetV2 architecture that extracts salient features from the input data that are then passed to an autoencoder to form more abstract representations prior to the classification layer. The proposed model also adopts extensive data augmentation techniques (e.g., rotation, flip, Gaussian blur, sharping, emboss, skew, and shear) to increase the number of samples for effective training. The performance of our proposed model is evaluated on three publicly available datasets and achieved the highest performance as compared to other state-of-the-art models.

RF Remote Blood Glucose Sensor and a Microfluidic Vascular Phantom for Sensor Validation.

Diabetes has become a major health problem in society. Invasive glucometers, although precise, only provide discrete measurements at specific times and are unsuitable for long-term monitoring due to the injuries caused on skin and the prohibitive cost of disposables. Remote, continuous, self-monitoring of blood sugar levels allows for active and better management of diabetics. In this work, we present a radio frequency (RF) sensor based on a stepped impedance resonator for remote blood glucose monitoring. When placed on top of a human hand, this RF interdigital sensor allows detection of variation in blood sugar levels by monitoring the changes in the dielectric constant of the material underneath. The designed stepped impedance resonator operates at 3.528 GHz with a Q factor of 1455. A microfluidic device structure that imitates the blood veins in the human hand was fabricated in PDMS to validate that the sensor can measure changes in glucose concentrations. To test the RF sensor, glucose solutions with concentrations ranging from 0 to 240 mg/dL were injected into the fluidic channels and placed underneath the RF sensor. The shifts in the resonance frequencies of the RF sensor were measured using a network analyzer via its S11 parameters. Based on the change in resonance frequencies, the sensitivity of the biosensor was found to be 264.2 kHz/mg·dL-1 and its LOD was calculated to be 29.89 mg/dL.

Abnormal Activity Recognition from Surveillance Videos Using Convolutional Neural Network.

Background and motivation: Every year, millions of Muslims worldwide come to Mecca to perform the Hajj. In order to maintain the security of the pilgrims, the Saudi government has installed about 5000 closed circuit television (CCTV) cameras to monitor crowd activity efficiently. PROBLEM: As a result, these cameras generate an enormous amount of visual data through manual or offline monitoring, requiring numerous human resources for efficient tracking. Therefore, there is an urgent need to develop an intelligent and automatic system in order to efficiently monitor crowds and identify abnormal activity. METHOD: The existing method is incapable of extracting discriminative features from surveillance videos as pre-trained weights of different architectures were used. This paper develops a lightweight approach for accurately identifying violent activity in surveillance environments. As the first step of the proposed framework, a lightweight CNN model is trained on our own pilgrim's dataset to detect pilgrims from the surveillance cameras. These preprocessed salient frames are passed to a lightweight CNN model for spatial features extraction in the second step. In the third step, a Long Short Term Memory network (LSTM) is developed to extract temporal features. Finally, in the last step, in the case of violent activity or accidents, the proposed system will generate an alarm in real time to inform law enforcement agencies to take appropriate action, thus helping to avoid accidents and stampedes. RESULTS: We have conducted multiple experiments on two publicly available violent activity datasets, such as Surveillance Fight and Hockey Fight datasets; our proposed model achieved accuracies of 81.05 and 98.00, respectively.