Crowdsourcing Solutions: GeoHIVE’s Unique Approach to Imagery Analysis

Up to 100,000 known features can be verified and assigned an attribution in GeoHIVE validation campaigns. This campaign asked the crowd to verify the presence or absence of airfields in the given AOI.

How can a crowd of different people from all walks of life provide accurate and timely imagery analysis? Welcome to GeoHIVE (Human Imagery Verification Effort), the crowdsourcing platform from Radiant Solutions that draws upon the wisdom of a vetted crowd to discover, validate, or define (create) features within an area of interest (AOI) to provide solutions for even the toughest geospatial problems. These problems could include marking all new construction sites in a specific city, validating the presence of airports in a region, or marking the level of damage to infrastructure after a natural disaster.

How does it work?

Individuals interested in joining the GeoHIVE crowd are provided guidance on how to approach imagery analysis then given a qualification test made up of imagery examples from live campaigns. Once they have passed the test, they are granted a qualification to work on closed campaigns.

Campaign imagery is pulled from the DigitalGlobe library or from third party sources, depending on availability and quality. Once the imagery is selected, it is chipped into smaller pieces, which are then distributed to the crowd. Each image chip in a campaign is seen by several crowd members to ensure all features of interest are found — this technique also allows for the crowd to verify each other’s answers. Once a campaign is complete, the images are then sent through a rigorous quality control (QC) check by GeoHIVE’s analysts to ensure the highest standard of results.

GeoHIVE’s crowd can identify over 100 features within 100,000 sq. km. in the course of 1 day. Here, they were asked to identify and tag any buildings found in the campaign images.

What are the types of GeoHIVE campaigns?

GeoHIVE campaigns are divided into three categories — discovery, validation, and editor (creation). In each, the crowd is requested to do the following:

Discovery — Find and tag features of interest that were previously unknown. These campaigns can ask the crowd to look for items such as vehicles, roads, bridges, new construction areas, etc. A recent discovery campaign was used to identify vehicle and foot crossings over the Venezuela-Colombia border. The crowd was able to identify 1,622 foot crossings and 1,590 vehicle crossings in an impressive 43 minutes.

Validation — Verify and label known objects within a polygon. These campaigns can ask the crowd to validate the presence of airfields, roads, buildings, etc. A validation campaign was most notably used to support emergency services in the aftermath of Hurricane Michael — crowd members were asked to assess damage caused by the storm by labelling the existing buildings as ‘No Damage Visible’, ‘Visibly Damaged’, or ‘Visibly Destroyed’. Using pre- and post-event imagery of the coastline from Mexico Beach to Port St. Joe, they assessed a total of 7,167 buildings within 3 hours and 8 minutes.

Editor — Digitize and label features of interest. These are unique campaigns in that the crowd is asked to find features of interest, draw bounding boxes around them, and then appropriately label them. An editor campaign was used in a program for the University of South Carolina where students were asked to digitize features in imagery dating from 1973 to 2017 along the South Carolina coastline. This campaign demonstrated the changes in the areas of interest over the 44-year timeframe.

Why the crowd?

There are many reasons the crowd proves to be a benefit in the realm of geospatial analytics, but one of the biggest is speed. GeoHIVE currently works with a crowd made up of more than 2,940 members. What a small team of analysts could do in a week or more, the crowd can tackle in a matter of hours. So, when time is critical, GeoHIVE can provide the results necessary to solve pressing issues. This is especially true in emergency response situations where speed is paramount to rescue effort success.

In 2018, GeoHIVE provided a damage assessment of the coastline east of where Hurricane Michael made landfall. In the Mexico Beach area, 2,117 buildings were assessed by the crowd using pre-event imagery from DigitalGlobe and post-event imagery from NOAA Rapid Response. They determined, of those buildings, 41% were visibly destroyed, 32% were visibly damaged, and 27% had no visible damage.

Thriving with the HIVE

Whether it’s aiding in humanitarian efforts, helping new businesses expand, or supporting security operations, GeoHIVE has the capabilities to deliver fast and accurate results for effective solutions to geospatial problems. If you’re interested in how you can join the GeoHIVE crowd and help out, feel free to visit geohive.radiantsolutions.com or email geohive@radiantsolutions.com to get started!