Convolutional neural networks (CNNs) are the gold standard in the machine learning (ML) community. As a result, most of the recent studies have relied on CNNs, which have achieved higher accuracies compared with traditional machine learning approaches. From prior research, we learned that multi-class image classification models can solve leaf disease identification problems, and multi-label image classification models can solve leaf disease quantification problems (severity analysis). Historically, maize leaf disease severity analysis or quantification has always relied on domain knowledge-that is, experts evaluate the images and train the CNN models based on their knowledge. Here, we propose a unique system that achieves the same objective while excluding input from specialists. This avoids bias and does not rely on a "human in the loop model" for disease quantification. The advantages of the proposed system are many. Notably, the conventional system of maize leaf disease quantification is labor intensive, time-consuming and prone to errors since it lacks standardized diagnosis guidelines. In this work, we present an approach to quantify maize leaf disease based on adaptive thresholding. The experimental work of our study is in three parts. First, we train a wide variety of well-known deep learning models for maize leaf disease classification, then we compare the performance of the deep learning models and finally extract the class activation heatmaps from the prediction layers of the CNN models. Second, we develop an adaptive thresholding technique that automatically extracts the regions of interest from the class activation maps without any prior knowledge. Lastly, we use these regions of interest to estimate image leaf disease severity. Experimental results show that transfer learning approaches can classify maize leaf diseases with up to 99% accuracy. With a high quantification accuracy, our proposed adaptive thresholding method for CNN class activation maps can be a valuable contribution to quantifying maize leaf diseases without relying on domain knowledge.

The demand for poultry products continues to increase in the world. The prevailing global climate change calls for improved poultry management systems to maintain optimal environmental conditions for boosting productivity. Increased food security based on sound, equitable and sustainable food production systems that utilize modern automation technologies is essential for all nations to be able to achieve the United Nations Sustainable Development Goals (UNSDGs). This proposed study seeks to help achieve some of the UNSDGs which include ending poverty, having zero hunger, good health and well-being, industry innovation and infrastructure. In this article, the design and implementation of a low-cost IoT-based remote poultry management system for small to medium scale producers is presented. Poultry farmers in developing countries are relying on manual poultry management methods which are labor-intensive. The proposed system which was built around the WiFi-enabled ESP8266 NodeMCU microcontroller is capable of monitoring and regulating temperature, humidity, water level, ammonia gas and the lighting system. Security is facilitated by the PIR sensor. The system minimizes employment costs and saves time. Besides, the system has unique capabilities of light scheduling and automatic switching control. The light schedules are pre-configured, and the user can select the required times of illumination. The illumination times in the evening are guided by the age of the birds. Light scheduling improves egg production and also conserves energy. The proposed light scheduling for the system was possible by executing a "cron job". A "cron job" enables the web server to perform repetitive and specific tasks at specific times. Remote monitoring and ease of accessibility of the system via the internet anywhere in the world using devices like smart phones and laptops is facilitated by our proposed web-based portal www.agrorun.co.zw/. The web page also allows users to turn ON or OFF the actuators like fans, blower fan/extractor fan, water pump and the lights.

A low-cost and light-weight ultra-wideband bowtie antenna for radar applications was simulated, fabricated and tested. A concise and easy to follow step-by-step description of the performed bowtie antenna simulation in Ansys HFSS software is presented. Optimized antenna parameters were utilized to fabricate the antenna. Fabrication was achieved by utilizing an FR4 PCB. The prototype was tested using a spectrum analyzer. The fabricated bowtie antenna results were used to validate the simulation results. The results obtained from the simulation platform were in close agreement to those of the prototype antenna. Results on effect of substrate thickness and frequency on S11 are also presented. The prototype produced improved overall S11 as compared to the simulation. The results indicate that the fabricated antenna satisfies bandwidth requirements for the UHF, L and S bands.

There is a continuous increase in health costs, thereby increasing pressure on individuals and consequently making the amounts claimed by the insured to be on the increase. In this study, data was collected from a large local insurance company in Zimbabwe for the period from January 2012 to December 2016. The aim of this study was to analyse the distribution and future pattern of insurance health claim system using time series approach. Akaike information criterion and Schwarz Bayesian criterion were used to select the adequate model through maximum likelihood estimation methods. ARIMA (0, 0, 0) (1, 0, 1) [12] is the model that was chosen to forecast claim amounts. The use of ARIMA models proves to be an excellent instrument for predicting and capturing the cost trend of health claims which can help in decision making to insurance companies. Keywords: ARIMA; Box-Jenkins; health insurance; time series.

In this paper, a detailed review of microcontroller unit (MCU)-based wireless sensor node platforms from recently published research articles is presented. Despite numerous research efforts in the fast-growing field of wireless sensor devices, energy consumption remains a challenge that limits the lifetime of wireless sensor networks (WSNs). The Internet-of-Things (IoT) technology utilizes WSNs for providing an efficient sensing and communication infrastructure. Thus, a comparison of the existing wireless sensor nodes is crucial. Of particular interest are the advances in the recent MCU-based wireless sensor node platforms, which have become diverse and fairly advanced in relation to the currently available commercial WSN platforms. The recent wireless sensor nodes are compared with commercially available motes. The commercially available motes are selected based on a number of criteria including popularity, published results, interesting characteristics and features. Of particular interest is to understand the trajectory of development of these devices and the technologies so as to inform the research and application directions. The comparison is mainly based on processing and memory specifications, communication capabilities, power supply and consumption, sensor support, potential applications, node programming and hardware security. This paper attempts to provide a clear picture of the progress being made towards the design of autonomous wireless sensor nodes to avoid redundancy in research by industry and academia. This paper is expected to assist developers of wireless sensor nodes to produce improved designs that outperform the existing motes. Besides, this paper will guide researchers and potential users to easily make the best choice of a mote that best suits their specific application scenarios. A discussion on the wireless sensor node platforms is provided, and challenges and future research directions are also outlined.

Appropriate design of intrusion detection systems is extremely important to safeguard critical and valuable infrastructure. Telecommunications companies (TELCOs) and electric power utilities are experiencing an increased rate of rampant underground copper cable theft and vandalism. This paper addresses the important subject of combating the physical intrusion of critical infrastructure by proposing a low-cost, effective and unique manhole intrusion detection system. Manholes are openings to confined underground spaces that are used to access critical underground infrastructure and utilities such as sewer systems, electricity and telecommunication networks for maintenance and inspection. Critical infrastructure protection is very essential to avoid losses and incapacitations that would have a debilitating effect on economic activities, security, public health and safety, and so on. The increasing rate of rampant copper theft and vandalism could be attributed to the increased demand and high prices for copper on the black market. Critical infrastructure sectors such as TELCOs, electricity supply, water and rail transport are utilizing intrusion detection systems that only trigger an alarm when the critical infrastructure has already been vandalized or stolen. In this paper, we propose a low-cost and unique manhole intrusion detection system with notification stages devoted to safeguard critical infrastructure. Simulations of the system were performed before hardware implementation. The proposed system utilizes an Arduino Uno microcontroller and multiple sensors to trigger the intrusion stages early before the copper cables are vandalized. Due to the inclusion of three different additional sensors, the proposed system has an advantage of timely or early-stage intrusion detection. The response time is improved since the first alert is sent early before the actual cutting of the cable. The GSM messaging system is utilized as the alert mechanism during the intrusion stages. Notifications are also displayed on a local LCD.

Wireless sensor network (WSN) technology has gained increasing importance in industrial automation, agriculture, smart cities, environmental monitoring, target tracking, structural health monitoring, healthcare, military applications, and so on. WSNs powered by batteries have a problem of limited lifetime due to energy constraints. Energy harvesting technology aims to eliminate the burden of replacing or replenishing depleted batteries for the sensor nodes by harnessing energy from the environment. Ultra-low power techniques are aimed at prolonging the overall sensor network lifetime by yielding significant energy savings in the WSN. The performance and lifetime of energy harvesting wireless sensor networks (EHWSNs) can be enhanced by the development of Dynamic Power Management techniques. Energy management and conservation are critical issues in EHWSNs, hence the need to develop energy harvesting-aware protocols and algorithms that facilitate perpetual network operation. It is anticipated that advancements in miniaturization and ultra-low power techniques will drive the widespread adoption of the energy harvesting paradigm. This article provides a comprehensive review of recent advances towards ultra-low power techniques in EHWSNs. We explore some of the existing types of power management techniques in WSNs including their disadvantages. The operating principles of recently proposed applications of ultra-low power techniques in EHWSNs are reviewed along with their associated fundamental mathematical expressions and assumptions. An analysis of these recent ultra-low power schemes is also presented. For each of the techniques, a summary of strengths, weaknesses and proposed solutions is presented. We provide the research community with open research issues and future research directions as well.

We have investigated the current-voltage (I-V) characteristics of nickel (Ni), cobalt (Co), tungsten (W) and palladium (Pd) Schottky contacts on n-type 4H-SiC in the 300–800K temperature range. Results extracted from I-V measurements of Schottky barrier diodes showed that barrier height (ФBo) and ideality factor (n) were strongly dependent on temperature. Schottky barrier heights for contacts of all the metals showed an increase with temperature between 300K and 800K. This was attributed to barrier inhomogeneities at the interface between the metal and the semiconductor, which resulted in a distribution of barrier heights at the interface. Ideality factors of Ni, Co and Pd decreased from 1.6 to 1.0 and for W the ideality factor decreased from 1.1 to 1.0 when the temperature was increased from 300K to 800K respectively. The device parameters were compared to assess advantages and disadvantages of the metals for envisaged applications.