Build models with multispectral imagery, map EUDR commodities from a point, & more

Welcome to GeoAI Horizons - your monthly look at the intersection of geospatial data & artificial intelligence!

In this edition of GeoAI Horizons, we look at how to build more accurate models with the help of multispectral imagery, how plot boundaries can be automatically mapped for EUDR commodities (such as coffee or cocoa) from just a single GPS coordinate, and more.

Go beyond what the human eye can see with multispectral models

Multispectral imagery refers to images captured at different wavelengths across the electromagnetic spectrum. Unlike regular RGB images, which capture light in three bands (red, green, and blue), multispectral images can capture light in multiple bands, often including infrared and ultraviolet. This allows for a more detailed analysis of the environment, as different materials reflect light uniquely across various wavelengths. Multispectral imagery can give greater insights into use cases involving crop health, soil conditions, environmental monitoring, land management, and more.

With geoAI software such as Picterra , you can build custom models on multispectral images to identify objects or patterns across different spectral bands that may not be visible or distinguishable to the human eye.

For example, by uploading a 4-band RGB+NIR image as a multispectral image, a custom model will use the information from the 4 bands, and the user can choose which bands to visualize and how to normalize each band. This choice is made by specifying what we call a bandspec. You can learn more about how to use multispectral imagery to build a model here.

Map plot boundaries for EUDR commodities from GPS coordinates

The EU Deforestation Regulation (EUDR) aims to stop imports of deforestation-linked commodities into the EU. It covers a wide range of products globally tied to deforestation. Operators must ensure that only deforestation-free items enter the EU, proving that they were not produced on deforested or degraded land post-December 31, 2020, and were legally sourced.?

Spatial information from Earth Observation (EO) technologies and Global Positioning Systems (GPS) is crucial in enabling operators to meet EUDR due diligence requirements. However, no single data set or map can offer the accuracy needed to serve as the sole source for determining a commodity’s compliance with EUDR.

What happens if operators only have a single GPS coordinate of a production plot rather than the boundary details? Simply placing a circle at commodity plot coordinates isn't enough to indicate actual boundaries and ensure accurate mapping for EUDR compliance. That's where innovations such as the 'bubble gum' process seen below can automatically map plot boundaries with only location and area details. Following the crop pattern, changes in canopy, roads, paths, and grasslands can be detected to guide the creation of accurate plot boundaries. Discover more innovations here.

Using change detection for visual investigations

Sweden's national broadcaster SVT used geoAI to analyze the growth of cemeteries in Ukraine since Russia's full-scale invasion. The investigative team used Picterra 's recently upgraded change detection workflow and satellite imagery from Planet to quickly and efficiently build an AI model. Read the story and watch the full video here.

Engage with the geoAI community

With summer coming to a close there are a wealth of upcoming events we will be attending where you can engage with other like minded space and geoAI professionals, including;

• World Space Business Week: September 16-20. The leading global business space event – brings together 1,600 decision makers under the same roof for a week of deal-making and intense discussions shaping the future of the industry. Catch Picterra 's pitch there on the 18th at 17:10!
• INTERGEO: September 24-26. A world-leading event for geodesy, geoinformation and land management, gathering together experts and visionary minds from all over the world.
• Google Cloud Summit Switzerland: September 26. Experience AI and cloud up close and look forward to exciting impulses, diverse contacts, and the new way to cloud.

That's it for this edition of GeoAI Horizons. Be sure to subscribe so you don't miss future editions! Do you have feedback or something you'd like us to cover? Don't hesitate to get in touch.