Detecting and Classifying Damage from Imagery

Assessing the impact of natural or man-made disasters on supply chain operations is vital, as elaborated in this workflow on monitoring risk to property insurers from cyclonic activity. Disruptions caused by unforeseen events at a particular node often have a ripple effect across the supply chain, amplifying the negative impact. The Thailand Floods of 2011 are routinely cited as a real-life example, where the automotive, electronics, and several other industries were heavily affected globally.

I recollect that it was difficult for me to procure a new computer hard drive in Mumbai in the aftermath of this incident—the product faced a supply crunch and the prices of existing stock were soaring. Therefore, to mitigate the impact of unforeseen circumstances (refer my Oil Spill, Forest Fire and Landslide detection posts), organizations are compelled to put in place rapid response systems that help reduce damage, provide relief, initiate backups, and ensure business continuity.

In this post, I will highlight two workflows involving damage assessment using imagery:

1. gauging the impact of the horrific 2020 Beirut Explosion

2. automated detection and classification of road cracks

   
   

The former was carried out by processing satellite imagery, while the latter involved deploying a deep learning algorithm on footage acquired from a vehicle dashcam.

---

Workflow 1 : Assessing Damage from the Explosion in Beirut in 2020

The Beirut explosion was one of the largest non-nuclear explosions to date. While heavy casualties were largely restricted to the vicinity of the blast site, damage to infrastructure and property extended tens of kilometers away. The intensity of the explosion was such that shockwaves were felt as far as Cyprus, located more than 250 kilometers away. As a result, Lebanon’s already fragile economy braced itself for a prolonged period of turmoil.

While the video demonstration of the satellite imagery analytics can be viewed separately, I will summarize the workflow here. Initially, I obtained two Synthetic Aperture Radar (SAR) satellite imagery datasets acquired prior to the explosion and estimated their coherence—a measure of how similar surface features are between the two datasets. The technical term for this entire procedure, culminating in coherence estimation, is Interferometric SAR (InSAR).

Subsequently, I estimated coherence between another pair of SAR datasets—one acquired before and the other after the explosion. Comparing these two coherence outputs helps reveal damage to buildings and other urban infrastructure. This is inferred from pixels exhibiting a noticeable change in coherence values, indicating that the underlying surface has undergone significant alteration.

While drastic shifts in coherence can also arise from other factors—such as agricultural harvesting—these can be systematically ruled out through contextual understanding of the terrain (Beirut being a dense urban center) and, if required, through ground-truthing.

Figure 1: Final Output of Damage Assessment of the Beirut Explosion (event date: 4 August 2020) using Interferometric SAR on Sentinel-1 satellite imagery datasets

Figure 1 also depicts the damage extent derived from analyzing two multispectral satellite imagery datasets acquired before and after the explosion. This process involved change detection using near-infrared reflectance. While Sentinel-1 SAR imagery is acquired through an active sensor that transmits microwave pulses, Sentinel-2 multispectral imagery is acquired passively by capturing reflected solar radiation.

However, this output (shown in Figure 2) is largely overshadowed by the InSAR-derived results, as the latter is more adept at detecting structural and geometric changes—a critical requirement for this workflow.

Figure 2: Areas in Beirut that suffered damage from the explosion (denoted by red pixels), derived from pre- and post-event Sentinel-2 multispectral imagery

Workflow Credits: ESA and RUS Copernicus

Do read this post if you would like a more detailed explanation of satellite imagery analytics.

---

Workflow 2 : Detecting and Classifying Road Cracks

I was first exposed to automated road surface investigation using deep learning at my firm’s GIS partner Esri's Developer Conference in Kolkata (2019). The demonstration—presented by Divyansh Jha, a data scientist at Esri—applied a deep learning algorithm to vehicle dashcam footage acquired along the Delhi–Faridabad highway. Divyansh indicated that the entire assessment process—from data acquisition to analysis and validation—took about a month.

Below is a video of a similar demonstration presented at another user conference:

Video 1: Road crack detection and classification using a deep learning algorithm on vehicle dashcam footage

As you would agree, studies of this nature—and their potential impact on road safety—are invaluable.

From both time and cost perspectives, geospatial analysis augmented by Artificial Intelligence, Machine Learning and Internet-of-Things can deliver substantial value. With the advent of 5G telecommunications, the adoption of such workflows is expected to become increasingly mainstream.

I have personally used Esri's ready-to-use deep learning Models to detect and classify building footprints and swimming pools from aerial imagery previously. The performance of these algorithms is impressive—the outputs genuinely left me mesmerized. Here is another video walkthrough demonstrating the use of Esri’s deep learning models to identify and classify powerlines.

---

ABOUT US - OPERATIONS MAPPING SOLUTIONS FOR ORGANIZATIONS

Intelloc Mapping Services, Kolkata | Mapmyops.com offers a suite of Mapping and Analytics solutions that seamlessly integrate with Operations Planning, Design, and Audit workflows. Our capabilities include — but are not limited to — Drone Services, Location Analytics & GIS Applications, Satellite Imagery Analytics, Supply Chain Network Design, Subsurface Mapping and Wastewater Treatment. Projects are executed pan-India, delivering actionable insights and operational efficiency across sectors.

My firm's services can be split into two categories - Geographic Mapping and Operations Mapping. Our range of offerings are listed in the infographic below-

Range of solutions that Intelloc Mapping Services (Mapmyops.com) offers

A majority of our Mapping for Operations-themed workflows (50+) can be accessed from this website's landing page. We respond well to documented queries/requirements. Demonstrations/PoC can be facilitated, on a paid-basis. Looking forward to being of service.

Regards,

Arpit Shah