| Abstrakt: |
Forest fires are still a huge problem in many countries because of the environmental, social, and economic damages affected. Applications connected to the prediction, management, and recognition of wildfires are enhanced recently. But, with utilize of technical solutions including the Internet of Things (IoT), artificial intelligence (AI), and wireless sensor networks (WSN), it can be feasible for developing early-warning methods with appropriate accuracy. With important sensors to measure variations in wind speed, temperature, and humidity, as well as for detecting the occurrence of fire and smoke, is a suitable manner for accomplishing this task, based on our view. Imaging sensors in WSN are utilized for collecting data in the target environment and deep learning (DL) approaches for forest fire detection (FFD) observe the attained images. With this motivation, this study develops a new DL model for forest fire detection named, the FFDNet technique. The presented FFDNet technique uses an emperor penguin optimizer with machine learning for forest fire detection in WSN. The goal of the FFDNet technique is to identify the occurrence of forest wire using sensors in WSN and DL models. Primarily, the sensor nodes transmit the images to the BS where the actual classification process takes place. In the presented FFDNet technique, the guided filtering (GF) technique is employed for the noise removal process. Besides, the presented FFDNet technique exploits a modified Xception network for a feature extraction process with root mean square propagation (RMSProp) optimizer. For the detection process, the kernel extreme learning machine (KELM) model is used in this study and the EPO algorithm can optimally choose its parameters. A wide range of experiments was performed by the FFDNet technique and the results are examined under diverse aspects. The simulation results reported the enhancements of the FFDNet technique over other DL models. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink