The performance of closed-loop tilt-control and adaptive-optics systems suffers when conditions change. Examples of changing conditions are angular extent of the object, signal-to-noise ratio, and characteristics of the disturbance. A simple learning algorithm motivated by neural network theory is developed to change the closed-loop gain in real-time to adapt quickly to changing conditions. This technique finds the correct loop gain within seconds with no operator intervention, which saves several minutes for each observation. Simulation and experimental results show improvement for both tilt-control and adaptive-optics systems.
Proc. SPIE. 9838, Sensors and Systems for Space Applications IX
KEYWORDS: Signal to noise ratio, Point spread functions, Telescopes, Detection and tracking algorithms, Data modeling, Computer simulations, Space telescopes, Palladium, Charge-coupled devices, Algorithm development, Probability theory, Binary data, Situational awareness sensors
The goal of this research eﬀort is to improve Space Domain Awareness (SDA) capabilities of current telescope systems through improved detection algorithms. Ground-based optical SDA telescopes are often spatially under-sampled, or aliased. This fact negatively impacts the detection performance of traditionally proposed binary and correlation-based detection algorithms. A Multiple Hypothesis Test (MHT) algorithm has been previously developed to mitigate the eﬀects of spatial aliasing. This is done by testing potential Resident Space Objects (RSOs) against several sub-pixel shifted Point Spread Functions (PSFs). A MHT has been shown to increase detection performance for the same false alarm rate. In this paper, the assumption of a priori probability used in a MHT algorithm is investigated. First, an analysis of the pixel decision space is completed to determine alternate hypothesis prior probabilities. These probabilities are then implemented into a MHT algorithm, and the algorithm is then tested against previous MHT algorithms using simulated RSO data. Results are reported with Receiver Operating Characteristic (ROC) curves and probability of detection, Pd, analysis.
Proc. SPIE. 9469, Sensors and Systems for Space Applications VIII
KEYWORDS: Point spread functions, Telescopes, Detection and tracking algorithms, Data modeling, Monte Carlo methods, Space telescopes, Atmospheric propagation, Atmospheric modeling, Electro optical modeling, Atmospheric optics
This research paper deals with methods for improving the performance of Electro-optical detection systems designed to find Resident Space Objects (RSOs). Some methods for detecting RSOs rely on accurate knowledge of the system Point Spread Function (PSF). The PSF is a function of the telescope optics, the atmosphere, and other factors including object intensity and noise present in the system. Due to the random photon arrival times, any observed data will contain Poisson noise. Assuming that other noise sources such as dark current and readout noise do not contribute significantly, the final source of intensity fluctuations in the data is the atmosphere. To quantify these fluctuations, an optical model of a telescope system is developed, and its PSF is simulated. In a long exposure image, the effects of the atmosphere are well characterized with the long exposure atmosphere Optical Transfer Function (OTF). In contrast, a short exposure image does not average the fluctuations as effectively. To model the atmosphere, random phase screens with Kolmogorov statistics are added to the optical model to observe PSF fluctuations in short exposure telescope data. The distribution of the peak intensity is analyzed for varying exposure times and atmospheric turbulence strengths. This distribution is combined with the Poisson random arrival times of photons to create a combined model for received data, which is then used to design a new detection algorithm. The performance of the new space object detection algorithm will be compared to a traditional algorithm using simulated telescope data.