N. G. Andrianov

The task of monitoring space objects from the ground is very important today. In Earth orbits are thousands of satellites, and even more debris. For all satellites must follow in order to adjust their orbit, if necessary, disposal of satellites from colliding with each other or with space debris. Effective and relatively inexpensive tools of surveillance and detection are the telescopes. With a series of frames taken by an interval of time necessary to detect moving objects in frames. Need image processing to highlight the marks. You can use the spatial-temporal regression, or first allocate a mark in the shot, and then work with the frame as a collection of point sources. Consider two image processing algorithm. Image processing requires a lot of computing resources. Scenes from the telescope come about every 5 seconds, so to handle a 10-megapixel shots in real time on ordinary desktop computers cannot. Use a special processor and expensive supercomputers, they take up much space and require specially written for them programs. The new models of processors are increasing the number of executing cores. To use all the computing resources of modern multi-core processors, the program runs multiple threads to process data. The most powerful is by far the graphics processor, which contains up to 512 cores. Using graphics processors, image processing algorithms can operate in real time, which is necessary for the detection of space objects using telescopes. The choice of processing algorithm is highly dependent on weather conditions and the number of stars. When cloud cover and a small number of stars in the field of view telescope, using complex spatial-temporal regression does not make sense, but in the opposite case, without such filtering is necessary.