Using AI and Computer Vision Tools to Enhance Parcel Detection from Right of Way Plan Sheets

Item #: 20230029

CONTACTS

Implementing Organization: ROW GIS
Champion: Ben Teran
Implementation Lead: Ben Teran
Development Team:
- Ben Teran (Lead, GIS Admin, UDOT ROW)
- Ryan Tiu (Member, Sr GIS Analyst, UDOT ROW)
- Steve Gourley (Developer, IT Master Engineer)
- Erik Neeman (Developer, IT Analyst, DTS UGRC)
- Nathan Kota (Member, IT Manager, DTS UGRC)

Article Written By: Ben Teran
Innovation Council Liaison: Marcus Volk
Innovation Team Coordinator: Winston Inoway

STATUS

Implementation Date: 5/01/2023
Adoption Status: Fully Implemented
Adoptability Note: Using AI to identify data on plan sheets might be useful for cataloging information on archived plans.

APPLIES TO

Topic: Data Collection, Analysis and Visualization
Organization(s): Asset Management, Central Construction, Central Maintenance/Facilities Management, Central Preconstruction, Data Technology & Analytics, Environmental Division, GIS, Planning, Railroad, Region 1, Region 2, Region 3, Region 4, Right of Way, Stormwater/Environmental
Job Role(s): Business Analyst, Construction Engineer, Information Specialist, Program Specialist, ROW / Permits
Tags: labor costs, labor productivity, rural development, urban development, inventions, tracking systems, information technology >> software, networks, artificial intelligence ( AI ), cartography, civil engineering, construction, investigation, project management, right of way, value capture

PROBLEM

The Right-of-Way (ROW) GIS group needed a way to identify parcels on thousands of plan sheets stored in ProjectWise. The previous process required that each document be reviewed, and if a parcel was found, it would be compared with existing information. It was expected to take years to complete the review and inventory work.

CHANGE

The GIS team worked with a programmer from the Department of Technology Services (DTS) to develop a program to process the plan sheet documents using Google Cloud and computer AI tools. Google’s Vision AI tool was used to search documents for the standard marking of a parcel number within a circle. Using Optical Character Recognition, the program extracted the parcel text to a spreadsheet along with other useful attributes of the original document.

RESULT

Using the tool to review 90,000 plan sheets resulted in a spreadsheet with around 56,000 identified parcels. The parcels will be grouped by county, reviewed, and added to the GIS parcel inventory. This information is a major leap forward toward the goal of creating a statewide parcel inventory. It helps identify all the unknown parcels and saves the ROW group years of work. This information will also benefit UDOT as a whole as it will greatly improve the inventory data that so many UDOT employees regularly use.