Selected work / 02 · Geospatial engineering
From map imagery
to usable geometry.
Scottish Title-Shape Extraction Pipeline: combining computer vision, spatial processing and validation for a high-volume title dataset.
The context
Finding the shape
inside the image.
The work involved extracting Scottish title shapes from WMS imagery and preparing them for spatial data workflows. Image-derived shapes needed checks before they could become database records.
At Nimbus Property Systems, I automated a title-shape pipeline for 11,000+ records using Python, OpenCV, PostGIS, WMS imagery and EPSG:27700 processing.
The workflow
Image processing meets spatial SQL.
Use WMS imagery
Work with the imagery associated with Scottish title records.
Extract candidate title shapes
Apply Python and OpenCV image processing, including colour and contour rules.
Process spatial geometry
Use EPSG:27700 processing and PostGIS as part of the geospatial workflow.
Validate before insertion
Apply area, topology, duplicate and failure-validation checks before database insertion.
Validation
Not every contour
is a usable record.
The pipeline applied multiple rules before database insertion, rather than treating every detected shape as a valid result.
My contribution
Automation with
quality controls.
- Automated the high-volume Scottish title-shape workflow using Python, OpenCV, PostGIS and WMS imagery.
- Applied EPSG:27700 processing and area, topology, colour, contour, duplicate and failure-validation rules before insertion.
- Diagnosed failed image extraction and geometry issues, using processing logs and QA tables to refine rules and improve reliability.
The delivery
A controlled path from image to record.
The resulting workflow joined image processing with spatial validation and database insertion. Its scope covered 11,000+ records, with checks designed to identify unsuitable or failed extractions.
Finding a boundary is one step. Making it a trustworthy spatial record is the engineering work.