Attitude Determination by Image Processing Algorithms

To determine attitude or direction with star constellations is perhaps the oldest method in the field of navigation and control. Besides its accuracy, star constellations can give three axis information by using just a single sensor, a digital CCD camera. What is very easy for men looking at the stars turns out to be complex if done by an onboard computer. The process of star identification can be divided into two parts: object segmentation and, needing much more calculation time, the matching process. Matching can be explained as comparing a star catalog stored in a computer memory with identified objects in a picture. A matching algorithm based on this principles has been developed and tested for the low cost satellite BREM-SAT. Tests have shown that at least four stars are necessary to identify a unambiguous attitude with 0.20 accuracy. Due to the comparably long computation time (20-120s), only the spin axis direction of the satellite is determined. This axis moves very slowly, but knowing its attitude is essential. Different test methods, including a hardware-in-the-Ioop test with the sensor matching simulated constellations on a computer screen, will be presented. For future projects, several improvements has been found which will reduce the computation time and give full three axis identification.