期刊
JOURNAL OF INTELLIGENT & FUZZY SYSTEMS
卷 43, 期 6, 页码 7081-7095出版社
IOS PRESS
DOI: 10.3233/JIFS-213308
关键词
Convolutional neural network; fake labeling; natural disaster; image classification
资金
- Deanship of Scientific Research at Najran University [NU/ESCI/17/049]
Disasters caused by natural events pose significant threats to human life and property. The influence of fake image posting on social media regarding natural disasters is increasing. Existing machine learning algorithms fail to identify fake labeling on disaster images and struggle to handle the classification process during multiple disaster events. To address this issue, a multi-model convolutional neural network (MMCNN) is proposed to accurately detect fake label images in multi-phormic natural disaster events.
Disasters occur due to naturally stirring events like earthquake, floods, tsunamis, storms hurricanes, wildfire, and other geologic measures. Social media fake image posting influence is increasing day by day regarding the natural disasters. A natural disaster can result in the death or destruction of property, as well as economic damage, the severity of which is determined by the resilience of the affected population and the infrastructure available. Many researchers applied different machine learning approaches to detect and classification of natural disaster types, but these algorithms fail to identify fake labelling occurs on disaster events images. Furthermore, when many natural disaster events occur at a time then these systems couldn't handle the classification process and fake labelling of images. Therefore, to tackle this problem I have proposed a FLIDND-MCN: Fake Label Image Detection of Natural Disaster types with Multi Model Convolutional Neural Network for multi-phormic natural disastrous events. The main purpose of this model is to provide accurate information regarding the multi-phormic natural disastrous events for emergency response decision making for a particular disaster. The proposed approach consists of multi models' convolutional neural network (MMCNN) architecture. The dataset used for this purpose is publicly available and consists of 4,428 images of different natural disaster events. The evaluation of proposed model is measured in the terms of different statistical values such as sensitivity, specificity, accuracy, precision, and f1-score. The proposed model shows the accuracy value of 0.93 percent for fake label disastrous images detection which is higher as compared to the already proposed state-of-the-art models.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据