top of page
  • Writer's pictureArpit Shah

Damage Detection & Classification from Imagery

Updated: Jun 13, 2023

Assessing the impact of natural or man-made disasters on operations is vital from a risk management and business continuity point of view. Prevention is always better - as elaborated in this hurricane risk case for insurance industry. However, several disasters go undetected and it is only after its occurrence does one opt for a rapid response and monitoring mechanism to reduce the damage, provide relief to those affected and conduct rescue operations, if necessary (as elaborated in oil spill, forest fire, seaweed use cases). The impact of a pause in operations at a particular node / location due to a disaster is often magnified at other nodes across the supply chain. The Thailand floods of 2011 is a perfect example - auto, electronics and several other global supply chains were heavily affected. I distinctly remember there was a phase where it was difficult to obtain a brand new computer hard drive in Mumbai in the aftermath of this incident! Ignore the 'bullwhip' effect of such situations at your own peril.


In this article, I will highlight two examples of assessing damage using imagery. The first is assessing the impact of the horrific explosion in the port city of Beirut, Lebanon in 2020 and the second involves detecting a phenomenon which is seemingly innocuous but very damaging in its own way - Road Cracks.


Both the damage assessments have been done using different ways - the former via analyzing optical and satellite imagery whereas the latter involves the use of GIS-based Deep Learning Framework on images captured from a vehicle's dashcam.

 

Example 1 : Assessing Damage caused by the Blast in Beirut

Final Output: Damage Assessment using Sentinel Satellite Imagery - Beirut Explosion - 4th August 2020
Figure 1: Final Output: Damage Assessment using Sentinel Platform's Satellite Imagery - Beirut Explosion - 4th August 2020

Without delving into the methodology of deriving this output, essentially I've taken two radar satellite images prior to the explosion (July 2020) and derived its coherence i.e. similarity / correlation between the pixels of the two images. Then, I derived the coherence of another two images - one image before the explosion and one after the explosion. The difference between the two coherence outputs is what is depicted in the image above.


Note: The image above also contains the damage extent as determined by analyzing optical satellite imagery - one image prior to and another post the explosion. However, it is largely masked by the damage extent as determined by analyzing the radar images. This is due to the fact that the properties of radar imagery make it more capable at capturing land surface changes information than optical imagery.

Damage Output (Red) as detected by analyzing pre and post explosion optical satellite imagery (Sentinel)
Figure 2: Damage Output (Red) as detected by analysing pre and post explosion optical satellite imagery

The Beirut explosion is one of the largest non-nuclear explosion till date. While loss of life was restricted to the vicinity of the explosion site, loss to property occurred tens of kilometers away. The shockwaves of the explosion was felt in adjacent countries (>250 kms).


With this tragic disaster, the already fledgling economy of Lebanon was put in further turmoil.


Much thanks to ESA & RUS Copernicus for the training material and imagery analysis methodology.

Do read this article if you'd like to know more about what satellite imagery analytics entails.

 

Example 2 : Detecting Road Cracks using Deep Learning


Do watch the demonstration of this amazing and effective geosolution from the video below -

Video 1: Road Crack Detection using DL demo video. Methodology can be found here.


The demo was conducted by my firm's GIS technology partner - Esri at its developer conference in 2019. I had the good fortune of seeing this demo live later at Esri User Conference in Kolkata in 2019. It was presented by Mr. Divyansh Jha - Data Scientist at Esri and the DL methodology was applied on Vehicle Dashcam images captured on the Delhi - Faridabad highway route. Basis my interaction with Divyansh - I gathered that the entire assessment process lasted a month - from capturing data to analyzing it and validating the results.


From both time and cost perspective - geotechnology blended with AI, ML, DL & IoT can be very beneficial. With the advent of 5G telecommunications, I do envisage these becoming mainstream and being able to solve a multitude of pressing problems faced by organizations and institutions worldwide.


I have personally used Esri DL models to detect new objects - buildings and swimming pools - from aerial images. The output was quite impressive - left me mesmerized. Also, you may refer my detailed video walkthrough on Using Esri DL framework to classify Power Lines here.

 

ABOUT US


Intelloc Mapping Services | Mapmyops is engaged in providing mapping solutions to organizations which facilitate operations improvement, planning & monitoring workflows. These include but are not limited to Supply Chain Design Consulting, Drone Solutions, Location Analytics & GIS Applications, Site Characterization, Remote Sensing, Security & Intelligence Infrastructure, & Polluted Water Treatment. Projects can be conducted pan-India and overseas.


Several demonstrations for these workflows are documented on our website. For your business requirements, reach out to us via email - projects@mapmyops.com or book a paid consultation (video meet) from the hyperlink placed at the footer of the website's landing page.

Regards,

249 views

Recent Posts

See All
bottom of page