Artificial Intelligence and IoT-Based Disaster Management System

Artificial Intelligence Based Disaster Management

Artificial intelligence (AI) is a system of cognitive processes that are similar to those of humans, such as learning new things, speaking, and solving problems. AI may also perceive its surroundings and behave accordingly to maximise its performance parameters or reach its objectives. Many nations, especially those that have had severe disasters or are prone to natural disasters, must implement a number of disaster management measures before, during, and after catastrophes due to the growing effect of natural disasters. Natural Disaster Management (NDM) is often separated into many stages based on the incidence, progression, and evolution of disasters. These stages include disaster preparedness, the time period prior to the catastrophe, disaster response, the time during and following the disaster, and disaster recovery, the extended period following the disaster. Figure 1 represents the Natural Disaster management in three phases.

Figure 1.

Phases of natural disaster management

The classification can be carried out using many decision tree algorithms, such as decision stump, hoeffiding tree, J48, Linear Model Tree (LMT), Random Forest, Random Tree, Representative (REP) Tree, J48 graft, and other well-known algorithms like LibSVM, Logistic regression, Multilayer perceptron, BayesNet, and Naive Bayes. Physical attributes including voice, scent, body warmth, motion, face form, skin colour, and body shape are employed to determine presence. Standard person detection algorithms and RGB picture databases may be used to quickly identify human bodies. Using a microphone, a cross-power spectrum approach is utilised to recognise voice. CO2 sensors measure gas emissions, and by identifying breathing patterns, they are able to identify human presence. Nevertheless, this technique has drawbacks due to its extended response time and the temperature, humidity, and dust content of the surrounding air.

• (i)Thermal and RGB pictures have been used in several studies to identify victims.

  • (ii)

    Pre-trained CNN architectures were frequently used in several convolutional layers. (In actuality, the majority of human detection techniques used concepts like bounding boxes and techniques like SSD, YOLO, and RCN, etc.; they are not in our field of interest.)

  • (iii)

    Accurate results were obtained quickly using machine learning techniques such as kNN, SVM, Naive Bayes, and Decision trees.

  • (iv)

    CNN architectures are commended for their enhanced feature extraction powers.