Building Footprint Detection Using Deep Learning Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing building detection systems are error-prone, require extensive manual labor, are slow, and not scalable, and are difficult to implement in urban and rural environments, leading to inefficient geospatial mapping and infrastructure planning.

Innovation Solution

An automated imaging system using a deep learning-based object segmentation method, such as a cascaded reverse mask RCNN framework, to extract building footprints from satellite imagery, enabling non-contact, non-destructive, and cost-effective 3D profile mapping of buildings and structures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional manual field visits and GPS marking are used to identify building footprints, then accuracy can be maintained through human expertise, but the process requires extensive manual labor, is slow, and is not scalable

Engineering Contradiction:
Improvebuilding footprint detection speedVSAvoidmanual labor requirement
Core Design Contradiction:
ProductivityVSExtent of automation

Solution Approach 1:

The patent replaces manual mechanical field visits and GPS marking with an automated imaging system that uses cameras to capture images and deep learning algorithms (specifically cascaded reverse mask RCNN framework) to automatically detect and segment building footprints, eliminating the need for human field agents and manual boundary marking

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service automation where the deep learning model automatically processes images to extract building footprints without human intervention. The cascaded reverse mask RCNN framework performs automatic segmentation and contour extraction, making the system self-sufficient in detecting and mapping building boundaries from imagery

Inventive Principle:
Principle #25Self-service

2Measurement precision

If specialized GIS experts manually label and curate building boundaries, then high precision can be achieved, but the process is extremely expensive and time-consuming

Engineering Contradiction:
Improvebuilding boundary accuracyVSAvoiddata curation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the mechanical process of manual GIS expert labeling and curation with an automated deep learning-based image segmentation system. The cascaded reverse mask RCNN framework automatically processes images to generate precise building boundary contours, eliminating the need for specialized human expertise while maintaining high accuracy

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs preliminary automated segmentation and contour extraction before any human review. The deep learning model pre-processes the imagery to identify building footprints and generate boundary contours, which can then be validated or adjusted if needed, significantly reducing the time and expertise required for final curation

Inventive Principle:
Principle #10Preliminary action

3Ease of manufacture

If conventional image processing methods are used for building detection, then the system is easier to implement, but the results are error-prone and not scalable to urban and rural environments

Engineering Contradiction:
Improvesystem implementation easeVSAvoiddetection accuracy
Core Design Contradiction:
Ease of manufactureVSReliability

Solution Approach 1:

The patent transforms the detection approach by changing key parameters: using deep learning architectures (cascaded reverse mask RCNN) instead of conventional image processing, and employing multiple processing stages with different parameters (segmentation thresholds, contour detection parameters) to improve reliability while maintaining scalability across diverse urban and rural environments

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4141789B1An automated imaging system for object footprint detection and a method thereof
Publication Date: 2025.11.19 JIO PLATFORMS LTD
  • EP4141789B1 patent drawingFigure 1
  • EP4141789B1 patent drawingFigure 2
  • EP4141789B1 patent drawingFigure 3

AI summary

The present disclosure provides for a system for facilitating a completely automated process that may directly fetch an imagery of a given location and area from any mapping module and extract a plurality of objects in the given imagery. Further, a deep learning-based object segmentation such as but not limited to a cascaded reverse mask RCNN framework method may generate a set of predefined vectors associated with the image. The system may be configured to automate the generation of the predefined vectors based on the image received from the image sensing assembly.