Our expertise
Raw imagery can bring various difficulties for most of EO application use cases. This is due to sensor imperfections, specific sensing conditions, flawed image geometry and geolocation, impractical data format and insufficient data description. Transforming raw sensor data and metadata to physically based, well-described and easily accessible EO products streamlines the EO mission value. This gets pronounced even more once the end-to-end or “raw-to-knowledge” processing is required.
We currently provide services for in-orbit calibration/validation and image processing towards readily-available EO products.
In-orbit calibration/validation
In-orbit calibration/validation (in-orbit cal/val) is a process of validating the ground calibration outputs (such as non-uniformity frames or geometric targets) and developing updated calibration outputs once the mission is on orbit. This part is a complex service for each mission or payload.
The early commissioning phase of the mission focused on sensing involves various analyses and trials, primarily focusing on real-world calibration, reference site imaging, and data-fit tests to identify errors caused, for example, by vibrations and temperature changes. Key activities include sensor failure detection, cross-device acceptance to determine biases between the camera’s optical axis and other devices, and vicarious calibration for radiometric and geometric corrections using natural or artificial sites. Additionally, cross-calibration compares camera images with those from established payloads like Sentinel-2 MSI. The continued in-orbit cal/val repeats these processes regularly and reacts to errors or changes, while minimizing disruptions to the satellite's nominal imaging. Continuous operations include procedures like dark-frame and flat-frame imaging, sensor monitoring, spectral response assessment, effective focal length and optical distortion monitoring.
Image processing
Image processing requires unprocessed imagery, up-to-date sensor calibration/validation data, acquisition metadata and some S/C metadata (e.g., original or derived information from other measurement components). Processing and cal/val respond to each other and are often part of the same process, however, here the processing refers to an application of the cal/val outputs and analytical image methods.
If we were to omit processing aimed to gather knowledge, such as model outputs or analyses, the processing operations generally deal with data errors, radiometric corrections, geometric corrections, time-space adjustments and product generation.
💥 Sensing errors as well as inaccuracies originating from processing are checked at the beginning and during the process and are either repaired or flagged. The whole acquisition and product might be assessed for expected use case eligibility if needed.
☀️ Radiometric correction adjusts for sensor and external errors related to energy transfers, transforming pixel values to accurately reflect the properties of sensed objects or surfaces. Raw pixel values are converted to Top-of-Atmosphere (TOA) radiance, TOA reflectance, and possibly to Bottom-of-Atmosphere (BOA) reflectance using atmospheric correction. Radiometric correction typically precedes geometric correction to avoid introducing errors. It involves operations such as flat-field correction, dark signal correction, radiometric error management, conversion to TOA reflectance, atmospheric correction, cloud masking.
🗺️ Geometric correction refines the sensor model to correct distortions and register the image accurately on Earth's surface. The process includes geometric model development and application, georeferencing, orthorectification, co-registration, image resampling and tiling, 3D structural modelling, stereophotogrammetry etc.
📦 Product generation encompasses creating analysis-ready products focused on the best possible accessibility. The processed images are accompanied with metadata, which are updated and consolidated throughout the processing pipeline. Additional assets may include quality masks, preview images, overview PDF/HTML, processing parameter datasets and similar.
Processing levels
Orbis is able to output Earth Observation products in different processing levels fitted for various observation and application purposes. The processing levels define properties (and possibly accuracies) of the resulting imagery, metadata and other assets. The acquisitions in Orbis can be processed to levels from Level 0 to Level 4:
| Processing level | Definition |
|---|---|
Level 0 (L0) | Consolidated and unprocessed instrument data in computer-readable formats at full time-space resolution. It contains metadata and other assets necessary for subsequent processing. |
Level 1A (L1A) | Radiometrically corrected data in sensor units. No geometric corrections applied. Metadata provide radiometric and geometric processing parameters for subsequent processing and relevant radiometric quality indicators. |
Level 1B (L1B) | Radiometrically corrected data in physical units (such as Top-of-Atmosphere reflectance). Geolocated data |
Level 1C (L1C) | Orthorectified data, resampled to a specific grid. |
Level 2 (L2) | Radiometrically processed or processable data to geophysical parameters (usually Bottom-of-Atmosphere reflectance, i.e., surface reflectance) and with consistent resolution and location as Level 1B data. |
Level 3 (L3) | Data or geophysical parameters that have been spatially, temporally resampled to certain degree of product consistency. |
Level 4 (L4) | Model output or results from analyses of lower level data. Variables are not directly measured by the instruments, but are derived from these measurements. It can be derived from multiple instrument measurements. |
The above specification shall be taken as a generic guide. EO products cannot are not equal between purposely different instruments/missions and thus, the processing levels are usually further detailed for each such instrument/mission. There is no barrier given to the EO product (re)specification as well as to its name.
Often, there can be variations between levels as a result of different requirements for radiometric and geometric processing. The EO product could be named Level 1 True Ortho if data are radiometrically processed to sensor units but the imagery is directly orthorectified. Similarly, the product could be named Basic Surface Reflectance Product if the exact geolocation or orthorectification is not important for the resulting product. For Level-3 and above, the name is dependent on the time-space alignment criteria or the performed analysis (e.g., L3 Pansharpened Product, 5-day NDVI, Number of cars).