1 of 10

Bridging the Gap: A Comparative Analysis of OSM and AI-Generated Data for Improved Disaster Response

Submitted by:

Pragya Pant (Mentee)

Supervisor:

Dr. Laxmi N Goparaju (Mentor)

​

May 2024

2 of 10

INTRODUCTION

  • Every year, the world is experiencing an increasing number of disasters (Meeting & Schuller, 2015).
  • Effective disaster response requires preparation and access to updated and complete data.
  • OSM (OpenStreetMap) databases and (Artificial Intelligence) AI-detected datasets are a quick way to access the data during the time of disasters.
  • OSM is an open-source platform for global land use and land cover data (Antoniou, 2017), whereas AI detection uses machine learning algorithms to automatically identify features in satellite imagery (Shi et al., 2020).
  • The damage caused by disasters can be assessed by analyzing pre- and post-disaster data extracted from AI as well as OSM data.

2

3 of 10

OBJECTIVES

The data obtained from OSM and AI generation may contain some data gaps. So, the objectives of this study are:

Primary Objective

  • To identify the most accurate and complete building data for disaster response.

Secondary Objectives

  • To calculate the total area and count of buildings from OSM and AI datasets.
  • To compare the accuracy and completeness of buildings from OSM and AI.

3

4 of 10

STUDY AREA

  • The Far-Western region of Nepal is prone to natural disasters, especially earthquakes (Sudhir, 2019).
  • The study area is the Bajhang district of Nepal which has been experiencing a greater number of seismic activities in the past few years.

4

5 of 10

METHODOLOGY

5

Figure 1: Workflow diagram

6 of 10

COMPARISON

6

Based on

OSM Data 2023

AI Detected Data 2023

Total Count (numbers)

58056

59925

Total Area (sq. km)

2578.21

2847.06

Table 1: Total count and area of buildings in Bajhang district

Figure 2: Map presenting OSM and AI generated building footprints of Bajhang district

7 of 10

VALIDATION

7

Figure 3: Sample study area showing OSM and AI-generated building footprints

8 of 10

RESULTS

  • qq

8

Figure 4: Graph showing the building count using different methods of detection

Figure 5: Graph showing building area for sample data

9 of 10

CONCLUSION

The result shows that, for speedy mapping, we can use AI-detected buildings but it needs more filtering and validation, whereas OSM data is validated by advanced mappers and can be relied on for local information but the data is not as complete as AI. Thus a combination of AI with OSM i.e. semi-automated system is recommended to make the mapping quicker and accurate improving disaster response.

9

10 of 10

REFERENCES

  • Antoniou, V. (2017). A Review of OpenStreetMap Data. In Mapping and the Citizen Sensor (pp. 37–59). https://doi.org/10.5334/bbf.c
  • Meeting, A., & Schuller, M. (2015). Disaster Response and Recovery: Aid and Social Change. 1–31. https://doi.org/10.1111/napa.12090.
  • Shi, W., Zhang, M., Zhang, R., Chen, S., & Zhan, Z. (2020). Change Detection Based on Artificial Intelligence: State-of-the-Art and Challenges. Remote Sensing, 145–151. https://doi.org/10.1145/3409501.3409532
  • Sudhir, R. (2019). SEISMIC HAZARD ANALYSIS OF NEPAL. June.

10