Scalable Methods for Brick Kiln Detection and Compliance Monitoring from Satellite Imagery: A Deployment Case Study in India

TL;DR

This work tackles scalable detection of brick kilns and regulatory compliance monitoring from satellite imagery in India, addressing the labor cost and geographic scalability of manual inventories. It deploys a scalable pipeline that samples public satellite imagery via Google Maps Static API, and leverages transfer learning from ImageNet together with self-supervised learning (SimCLR and JigSaw) to train robust detectors across geography shifts. Key findings include the identification of 7477 new brick kilns across 28 districts in 5 states within a 276{,}000 km^2 region, and a finding that 90% of Delhi-NCR kilns violate the mandatory 1 km separation, with 7.65 million people within 1 km of a kiln. The authors also release a web application for interactive kiln detection and demonstrate a reproducible pipeline with potential global applicability for inventory management and policy enforcement.

Abstract

Air pollution kills 7 million people annually. Brick manufacturing industry is the second largest consumer of coal contributing to 8%-14% of air pollution in Indo-Gangetic plain (highly populated tract of land in the Indian subcontinent). As brick kilns are an unorganized sector and present in large numbers, detecting policy violations such as distance from habitat is non-trivial. Air quality and other domain experts rely on manual human annotation to maintain brick kiln inventory. Previous work used computer vision based machine learning methods to detect brick kilns from satellite imagery but they are limited to certain geographies and labeling the data is laborious. In this paper, we propose a framework to deploy a scalable brick kiln detection system for large countries such as India and identify 7477 new brick kilns from 28 districts in 5 states in the Indo-Gangetic plain. We then showcase efficient ways to check policy violations such as high spatial density of kilns and abnormal increase over time in a region. We show that 90% of brick kilns in Delhi-NCR violate a density-based policy. Our framework can be directly adopted by the governments across the world to automate the policy regulations around brick kilns.

Figures (7)

• Figure 1: https://brick-kilns-detector.streamlit.app/ to detect brick kilns in a) user-specified bounding box; b) predicted brick kilns by our model; c) GradCAM selvaraju2016grad run to validate that our model actually focuses on brick kiln to make a decision; d) 684 out of 762 brick kilns (90%) in a highly populated area called Delhi-NCR are within 1 km distance from another kiln, which is a violation of the government policy (violating distances are shown in red color).
• Figure 2: Our brick kiln (labeled) datasets from (a) Bangladesh and (b) Delhi-NCR.
• Figure 3: Satellite images of a sample brick kiln from (a) Bangladesh and (b) India.
• Figure 4: Brick kiln labeling App. Labeling individual images can be time consuming due to sparsity of brick kilns, thus, we develop an App to label a batch of 25 images ($5 \times 5$ grid) at once.
• Figure 5: Brick kilns violating government policy of minimum distance 1 km between two brick kilns. 684 out of 762 (90%) brick kilns violate this policy. Red dots show the kilns violating the policy and black dots show the kilns within the regulations.