Not all satellite imagery and spatial intelligence is ready to be analyzed by AI/Ml algorithms. In fact, there's a good amount of work that needs to go into ensuring that the imagery is AI-rady.

There are three key requirements for satellite imagery to be considered analysis-ready, and they serve as the foundation for how Vantor prepares it spatial foundation for AI-powered analytics:

1. Atmospheric and radiometric correction
2. Orthorectification and alignment
3. Accelerated access and direct delivery

Atmospheric and radiometric correction

Vantor has a patented Atmospheric Compensation technology for satellite imagery analytics, which is critically important for effective analysis. Without compensating for atmospheric effects or calibrating the spectral bands, it is more difficult to correctly distinguish between relevant and nonrelevant features in the imagery or to rapidly analyze images across time and space.

The industry currently uses multiple processes for handling these corrections, and it is easy to inadvertently analyze a time-series stack of images with wildly varying colors and consistency. This often leads to incorrect results and the need to develop location-specific artificial intelligence and machine learning (AI/ML) workflows. Vantor eliminates this problem by providing consistent inputs ready for you to analyze immediately upon delivery.

We also apply our Dynamic Range Adjustment process to the pansharpened RGB imagery to help further color-balance the imagery and enable rapid identification of objects with unique color signatures or those in more challenging locales, such as under trees or along the edge of an image.

Orthorectification and alignment

Orthorectification and alignment are another important procesing step to ensure that the algorithm understands which features in the imagery are relevant. Orthorectification ties images to the correct location on Earth, and alignment controls for shifts between roads, buildings, and other features within a stack of images taken from different angles.

Orthorectification and alignment provide higher accuracy when detecting change, extracting features or monitoring transient objects, such as cars, ships and planes. These processes reduce false negatives and false positives from automated exploitation workflows. Misaligned or poorly orthorectified imagery can cause you to detect two buildings where only one exists or instruct a driver to make an off-road turn onto the sidewalk.

We apply our Bundle Block Adjustment (BBA) process to further refine the alignment of images within the ordered stack. BBA can be deployed on up to 50 images to create a deep temporal stack of aligned imagery. The deeper stack combines signals to provide further confidence in automated object detections and allows you to more accurately analyze changes over time.

Accelerated access and direct delivery

Imagery can be perfectly aligned, calibrated and optimized for a cloud-based workflow, but all those benefits are lost when you are forced to use cumbersome ordering and delivery processes. As with the requirement to optimize for cloud computing, data must be quickly discoverable and orderable to be “analysis-ready.”

APIs can enable faster analysis of satellite imagery through automation, batch ordering and machine-to-machine communication. Direct delivery into your compute environment speeds up your analysis because you don’t have to wait for a download website to be stood up or switch between systems to pull in your order.