How to segment and extract features from high resolution Capella Space SAR imagery

Questions & Answers

Yes, the dataset used for segmentation and object extraction is Single Look Complex (SLC) imagery

We are working closely with the Capella team to bring InSAR capabilities for Capella imagery into CATALYST

We use machine learning regression analysis. The main algorithms are Support Vector Machine and Random Forest. The the algorithm automatically learns from the training dataset, so as more training datasets are provided the better the algorithm learns

Capella’s standard SAR data imagery products are StripMap @ 1.2m and 5km x 20km, Site @ 1.0m and 5km x 10km, and Spot @ 50cm and 5km x 5km

The Capella Console is our automated tasking and SAR imagery browsing platform used today by government and commercial customers. If you would like to request a demo and access for your organization, you tell us more about your use case at and one of our account executives will find a solution that’s right for you. The console also contains free data sets, available through our community program. We are working with academic researchers, software engineers and disaster response groups through our community today and encourage you to register here:

Capella currently offers HH polarized X band SAR. As we gather additional customer requirements for their missions and use cases we will evaluate the addition of VV (current systems) or cross-polarized sensors (future spacecraft systems).

Official geolocation accuracy metrics are actively being characterized by Capella Space and will be published soon. The duty cycle highly dependent on the number and duration of collects taken in an orbit, and ranges from 6 minutes in a 90 minute orbit to 10 minutes in a 90 minute orbit.

Capella has supported glaciology and search and rescue missions related to the movement of Arctic and Antarctic ice sheets and icebergs. We would be happy to discuss specific use cases and make recommendations.

The noise in SAR imagery, typically referred to as Noise Equivalent Sigma Naught (NESZ), varies with look angle and range resolution. Capella Space is actively characterizing official NESZ metrics which will be available soon.

There are many applicaitons where having multiple polarization options can be a useful tool, crop classification and changes in surface materials are two examples. Our current constellation offers HH polarization and has the potential to capture using VV polarization.

Capella does not have a DEM product at the moment, but we expect that a customer would be able to generate a DEM form Capella data at meter-scale accuracy.

It is possible to detect multiple classes, and in some cases the type, of objects such as aircraft and vessels from high resolution SAR imagery. At Capella we are currently using machine learning to extract information from our very high-resolution SAR imagery to address customer needs.

The image resolution is 30 cm in the slant plane. The ground resolution will depend on the look angle of the collect. The azimuth resolution that Capella can deliver is limited to 50 cm by NOAA regulations.

Capella’s satellites are approximately 10 times smaller and lighter compared to previous generations of SAR satellites. The advantage of a small satellite design is that we are able to rapidly deploy a constellation of very high resolution sensors and iterate our technology. With a growing constellation covering multiple inclinations our customers benefit from a large variety and capacity of imaging opportunities precisely when and where they require information – all accessible via a web-based secure and private automated platform.

The standard off-nadir angle range is 25 deg. to 40 deg. Capella also offers extended and custom options with the ability to task our satellites upto a range 5 deg. to 45 deg.