| Title | Smart Lightning Detection System for Smart-City Infrastructure Using Artificial Neural Network |
|---|---|
| ID_Doc | 38509 |
| Authors | Ullah, I; Baharom, MNR; Ahmad, H; Wahid, F; Luqman, HM; Zainal, Z; Das, B |
| Title | Smart Lightning Detection System for Smart-City Infrastructure Using Artificial Neural Network |
| Year | 2019 |
| Published | Wireless Personal Communications, 106, 4 |
| Abstract | Smart city infrastructure for lightning detection is one of the most important parameters for building protection. To get outcomes within a short frame of time having high accuracy Artificial Neural Network (ANN) is a better choice to be used. In this work, ANN is applied for automatic building detection process for the application of smart-city infrastructure, which has drawn very little attention of researchers due to unavailability of standard data sets and well-defined approach. Object detection follows the lightning strike pattern of the lightning flashes on different air terminals installed on multi-geometrical scaled structures. Initially, classification is carried out based on the object characteristics into different categories. In the proposed approach, the classification of buildings has been carried out on the basis of various states of terminals of different buildings. The proposed approach consists of four stages namely data collection, data labeling, classification; and performance evaluation. In the data collection stage, data is collected from different scaled buildings by switching on and off states of different terminals. In the data labeling stage, the data collected are given labels according to the types of buildings. The buildings have been categorized on the based on lightning air terminals installed on it. In the classification stage, ANN with different combinations of network training function, hidden layer transfer function output layer transfer function, number of neurons in the hidden layer and different number of epochs has been used to classify the buildings into their respective classes. Difference performance and accuracy was found for the evaluation of the work and the highest accuracy was found to be 92.6 followed by 85.27, 84.82, 82.81, 81.72, 80.18 and 79.75 for various architectures of the network. For the validation of the methodology, other types of classifiers have also been applied for the discrimination of different categories of the buildings. |
No similar articles found.