Star Sensor and Algorithm Developments

1. History and achievements
2. Downloadable papers
3. Spin-off successes
4. Tool features and availability
5. Algorithms, performance, future developments
6. Example of tool's user interface

History and achievements
In February 1998, Delta-Utec developed and successfully tested the world's first experimental star sensor algorithms based on neural networks, or in fact, the so-called Parallel Distributed Processing, or PDP, allowing super-simple data input such as the number of stars and a histogram of brightnesses. Since then we have at regular interval developed new algorithms for attitude determination, simulation, validation etc.

Another algorithm, the optimized Douma triangle algorithm (DUDE) triggered the development of SSATT, a Star Sensor and Algorithm Test Tool, for sensitivity analysis of star sensor algorithms and camera design parameters. It supports the design and understanding of cameras as well as of algorithms. It includes a fully developed set of revolutionary algorithms. The SSATT was initially developed for ESTEC. SSATT can produce in batch its own simulated centroided images and perform automated attitude determination and statistic analysis. Today there are over 60 registered SSATT users worldwide. SSATT has been extended under internal and Dutch government funding to the ADTE: Attitude Determination Test Environment.

The ADTE is used for assessment and optimization of detection and centroiding algorithms using real camera imagery. It can process in batch raw images and/or precentroided images. (Semi)automated statistical analysis is performed and accuracy in arcseconds can be determined even without availability of reference measurements. (Semi)automated correction of consistent image deformations by lens imperfections is performed using genetic algorithm optimization, generally improving performance by a factor of 2.

A successful application in 2000 has been assessment of TNO-TPD's MEFIST II camera performance in the Israelian desert. Attitude has been determined for hundreds of images at several arcsecond accuracy. Several spin-off ideas are under development.

An embedded software version of the algorithms has been developed in C++ and is commercially available.

In 2008 the algorithms are being revived and upgraded for a new multiple aperture camera. Tracking is now included.

Select Star Sensor Technical Paper for Download

---

Spin-off successes!
- The extremely fast database search algorithms that were developed for SSATT were successfully applied in Orthopedic Specialty Delivery software, reducing database search times with a factor 100 (!), thus significantly improving the company's edge with respect to its competitors.

The SSATT/ADTE features:

- Variable CCD camera parameters;
- Variable algorithm settings;
- Statistical investigation tool;
- Automated on-board database generation;
- Monte Carlo simulations and automated report.
- Centroiding simulation at various levels of reality possible.
- Attach a real camera or batch process raw digital images;
- Optimization of adjustment for lens imperfections.
- Assess practically obtainable precision of your star sensor even without reference measurements.
- Batch investigate recognition and magnitude reconstruction of individual stars.

- SSATT demo-tool freely available (export restrictions apply)
- C++ embedded software version
- ADTE and/or tech support to your project commercially available

The performance of our algorithms:

- Attitude determination at arcsec accuracy;
- 99.99% reliability;
- On-board database size tens of kilobytes;
- Unprecedented identification and validation of stars: 95% per image.

The algorithms include:

- Triangle recognition methods:

. Douma,
. DUDE (Delta-Utec Douma Extension),
. Liebe,
. Quine;

- QUEST attitude determination;
- Automated prediction, correction and recognition of apparent multiple stars.
- Camera magnitude prediction models, image magnitude reconstruction
- Ultra-fast database search-by-pointer;
- Automated database uniformization;
- New revolutionary validation algorithms such as:

. Validation by doubles,
. Snowball validation: fast reliable validation without number crunching.

Under development:

- Automated attitude propagation;
- Rate determination.






To Home