Information
AI Chat

Chu Jun Rong TP060744 - Data Management Individual Assignment (Research)

cool

Course

Data Management

5 Documents

Students shared 5 documents in this course

University

Asia Pacific University of Technology and Innovation

Academic year: 2022/2023

Uploaded by:

Anonymous Student

This document has been uploaded by a student, just like you, who decided to remain anonymous.

Asia Pacific University of Technology and Innovation

Comments

Please sign in or register to post comments.

Related Studylists

Year2_Assignment

Preview text

1 Introduction
- 1 Data Mining
- 1 Flood Detection, Rescue and Evacuation................................................................................
2 Motivation for Using Data Mining in the Area
3 Dataset and Data Type Used
3 A Hybrid Model of Sectorization & Evacuation Path Detection for Disaster Affected Areas
3 Application of Deep Learning on UAV-Based Aerial Images for Flood Detection - 3 Emergency Shelter Geospatial Location Optimization for Flood Disaster Condition: A Review
3 Guidelines for the Use of Unmanned Aerial Systems in Flood Emergency Response
Technologies and UAVs: A Review 3 An Integrated Approach for Post-Disaster Flood Management via the Use of Cutting-Edge
4 Best Data Pre-Processing and Transformation Methods
- 4 Convolution Neural Networks (CNN)
- 4 Linear Programming
5 Data Mining/ Machine Learning Techniques Used
- 5 Image Classification..............................................................................................................
- 5 K-Means Clustering
6 New/ Prospective Opportunities Product by the Analysis
- 6 Unmanned Aerial Vehicles/System (UAVs)
- 6 Flood Warning System..........................................................................................................
- 6 Digital Elevation Model (DEMs)
7 Challenges Face in Flood Detection, Rescue and Evacuation
8 Future Directions of Research in the Flood Detection
9 Conclusion
References

2 Motivation for Using Data Mining in the Area

According to the Research done by Centre for Research on the Epidemiology of Disasters CRED, there are a high number of catastrophic events recorded with 432 cases and the flood disasters have occupied with 223 occurrences. Also, In China, the Henan Flood has taken away over 350 people lives and affects the China’s economy with 16 billion US dollars. This has brought a high impact to every country that are facing flood disaster each year. Also, the government of the countries may face funding problems on providing facilities to overcome this flood issues. Hence, these confirmed flood cases have brought motivation in conducting this research for having the prevention on flood detection, rescue and evacuation mission to prevent the loss of lives and impact to economy. According to the journal on Guidelines for the Use of Unmanned Aerial Systems in Flood Emergency Response, the researchers has conducted an online survey on getting the participants responses for creating the flood risk management activities model using the Data Mining Techniques. Besides, in the second journal which are Hybrid Mode of Sectorization & Evacuation Path Detection for Disaster Affected Area, the Flood Disaster has highly affected the facilities and environment of the countries which leads to loss of shelter, food and beverage shortages, and spread of the unknown disease which came from the flood in 2019. As a result, research and analysis on Flood Detection should be performed to ensure that the citizens and the country could have a prevention and a secure way on handling the flood disasters.

3 Dataset and Data Type Used

3 A Hybrid Model of Sectorization & Evacuation Path Detection for Disaster Affected Areas

Figure 1 Pre and Post Disaster Image of Mexico City Table 2 Disaster Level Table The dataset that has been used in this journal articles is the dataset form xBD Dataset repository which is about the Satellite Image Dataset of Natural Disaster Affected Area that Includes Flood Disaster. The dataset that has been collected contains the satellite images of different types of disasters which mainly focusing on flood disasters. All images are annotated that contains building polygon co-ordinate, classification of disaster affected area with intensity of damage and satellite metadata for building models. The satellite image is annotated as polygons using ISO/IEC 13249-3:2016 standard (‘ISO/IEC 13249-3’, 2016). From the Table 1, it shows the joint damage scale used in the dataset for classification of disaster affected area. The dataset from the figure above, it shows the Pre and Post disaster image which damaged the city in Mexico.

3 Emergency Shelter Geospatial Location Optimization for Flood

Disaster Condition: A Review

Figure 3 Overall Process on Article Pre-Processing

In this journal article, the researchers aim to collect the existing methodologies from

the Article Finding Process and discover for a better approach for the Flood Evacuation

Problem. The related articles and datasets were collected by using the Google Scholar and the

Open Knowledge Map by finding the keywords on “Evacuation shelter Optimization” and

“Emergency Shelter Optimization” with 938 references and 34 compatible articles

respectively. Afterwards, those articles are then collected to proceed for Article Screening to

choose the latest citation year within 2009-2020 which are suitable for this Flood Detection,

Rescue and Evacuation research with 26 articles evaluated.

3 Guidelines for the Use of Unmanned Aerial Systems in Flood Emergency Response

Emergency Response

Figure 4 3x3 matrix

In this journal article, the researchers have conducted some Literature Review and Face-

to-Face Interviews towards applicants in England and India. Those datasets will then be

collected as additional information of the Main Unmanned Aerial System (UAS) applications

and potential usage. The researchers have used 3 x 3 matrix to identify the potential usage of

UAS in Flood Emergency Response and Risk Management activities. Meanwhile, a number

of key factors were considered such as catchment size, flood source type and phase of flood

disaster for the x and y axes when deciding for the UAS flight plan deployment. Also, the

dataset that will be used in this journal articles are then to be converting into a Catchment Flood

Response Model to developing into the UAS deployment.

4 Best Data Pre-Processing and Transformation Methods

As from all the journal articles above, most of the articles are prepared based on the Unmanned Aerial Systems which the datasets are mainly in Aerial or Satellite Images data. Before starting with Data Pre-Processing actions, the suitable ways and methods to perform the data pre-processing will be selected through the journal articles. 4 Convolution Neural Networks (CNN) Figure 6 Convolution Neural Networks (CNN) explanation Convolution Neural Networks (CNN) is a machine learning that can used for neural network sketched for pre-processing the images into structured arrays of data which are suitable for computer vision as shown as the Figure above. (GeeksforGeeks, 2022) In this scenario, after the data collection, the researchers can perform CNN for capturing the data from the cropped image and patching on labelling the flood detection in the entire training procedure. 4 Linear Programming Figure 7 Flood Detection Algorithm Linear Programming is a method that provides practical features on solving the disaster to perform decision-making during the disaster happened. It will be the suitable way to solve the quantitative correlation problem. Based on the Figure provided, this will be the flow to used for the data which processed with Linear Programming. It can separated with Flood or Non Flood area with the Prediction Maps.

5 Data Mining/ Machine Learning Techniques Used

5 Image Classification Figure 8 Image Classification As Convolution Neural Network (CNN) was used for the Data Pre-Processing Method, it was used on differentiate the signal and image processing problems such as the texture and edges from an image into flooded and non-flooded categories. Also, it can be used for identifying the victim’s identification and the travel distance grouping. For doing classification, the datasets have been correctly identified the roads and buildings with 91% and 94% accuracy respectively. Figure 9 CNN Model Accuracy Result As classification is having a high percentage of accuracy, the researchers decided to start training the CNN model. The dataset will be thrown into the training session, and it results in 728 out of 800 images being correctly classified. Based on the table above, it shows the accuracy of the altered dataset that has been pre-processed with 91% accuracy.

6 New/ Prospective Opportunities Product by the Analysis

6 Unmanned Aerial Vehicles/System (UAVs) Unmanned Aerial Vehicles/System (UAVs) which is known as Remote Sensing Technologies is helpful in capturing images. UAVs are able to capture high-quality aerial photographs, aerial films, and vast amounts of precise data because to their high-resolution cameras and superior sensors. For a thorough investigation, the collected data is turned into granular 3D maps and models. In catastrophe zones, 3D mapping is very useful for revealing damages, flaws, or other dangerous aspects. For this scenario, the government could develop more drone system focusing more on the flood rescue mission such as detecting or searching for victims with Thermal Imaging features capturing within a big area in order to search in a smaller area or riding boat to a specific area only. 6 Flood Warning System The Flood Warning System could be developed to the disaster area with high frequency on flood disaster. Flood Warning System would send the warning signal to the nearest branches to warn the citizens on evacuate to a higher and safer place. Other than that, it would send the warning signal to the government just in case of any flood situation to perform rescue mission. 6 Digital Elevation Model (DEMs) Digital Elevation Model (DEMs) is a 3D computer generated graphics that represent the elevation data of the ground surface. It can be considered as the primary topographic input data for flood simulation and mapping for the future flood detection and its evacuation way.

7 Challenges Face in Flood Detection, Rescue and Evacuation

During conducting research and review, there are several issues that are being faced which is the different type of datasets requires data pre-processing on changing the raw images or removing the unwanted background from the Aerial Images dataset. As the datasets is connected with the Satellite and the Aerial System, most of the images are in a low resolution quality as these system could not generates better quality images which may affect the accuracy and the outcome of the pre-processed and tested data. Based on these problems, it requires data deletion for removing the unwanted background or obstacles by using formulas within it. Besides, the images requires data mining techniques for abstracting the classification from the images and turns it into an informative parts. This has leads to problems of lack of understanding towards the data mining techniques which requires doing more researches before starting using it to generalise the data.

8 Future Directions of Research in the Flood Detection

In the future, flood detection could be the first and foremost issue to be consider. Hence, the proposed system could have included with more advanced data mining techniques instead of Unmanned Aerial Vehicles/System (UAVs), Flood Warning System, and Digital Elevation Model. Also, the researchers could consider on implementing Artificial Intelligence (AI) so that all of the provided detection or evacuation would have the thought same as the human thinking.

9 Conclusion

Flood disaster is the most headaches problems that every country is facing as it may affect the whole country’s economic and leads to unknown illness. Thus, to prevent the flood disaster, data mining has an important role on helping the country on flood detection, rescue and the evacuation mission that may save citizens life. Data mining techniques especially in flood detection could predict the best evacuation roads and the nearest emergency shelter to avoid the flood disaster. All in all, flood detection could be developed with provided if the Predictive Model is having a high accuracy percentage which all researchers should maintain it with 99% so that the citizens could believe with the outcomes.

Was this document helpful?