Asteroid Hunter: Identification Using Matching Probability

ASTEROID HUNTER Star Identification

Using Matching Probability

Destin Smith-Norris CAP4401 2009F [email protected] November 23, 2009

Objectives • To accurately identify individual stars in an image • To use these stars to detect objects which are not recognizable • Do this using only the basic information able to be calculated - the stars' relative brightness, relative location to other stars, and number of near neighbors

Relevance • For my project, to attempt to locate asteroids in an amateur astrophotopher's image • In the real world, star identification is critical for space based devices to calculate their attitude, or location, in Space • Without accurate positioning in Space it would be impossible to keep satellites in Orbit

Algorithm Used • There are many popular starID algorithms used in Space, however there was one recently published by Xie Jungeng, Jiang Wanshou and Gong Jianya using matching probability

• The success of my application depends on three assumptions

1)that there will be enough nearby stars to compare against, 2)that the brightest neighbor to the base star will be the same in the user image and the database 3)that there is minimal distortion within the image

Algorithm 1.The algorithm as implemented takes both the database image and the user supplied image and creates for each a matrix of all the star locations sorted by brightness 2.Then, for each star the program finds all the neighbors and calculates the angular distances by pixel and stores this along with the location 3.Finally, the ratio of the angular distance of each base star to every one of its neighbors is calculated using the next brightest neighbor as a

Algorithm Cont…

Algorithm Cont… • The next step is to loop through every star in the user image database and compare the ratios of each base star to every star in the database • When two ratios are almost the same, they are considered a match. When two stars have enough matches, they are considered to be the same star and it’s coordinates are recorded

Results • Using my sample images results in many false star identification. This is because, I have determined, the fact that my camera is uncalibratedto the database • This means that the ratios do not match up accurately • To test this theory, I used my application to test matching a subsection of an image to the whole and it succeeds in accurately identifying the stars

Original User Image

Matches in Full User Image

Matches in Cropped Image

Results • Using a threshold of .001, a minimum of 6 probable matches, and a neighborhood of 100 pixels, I got these results. • 27 identified stars, out of 103 stars in the cropped image. • Using this user supplied image against a database image extracted from Stellarium gives at most a few correctly identified stars.

Conclusion • I believe if used with calibrated equipment and a database designed for the particular camera, this algorithm would be completely feasible for use in a real-world mission critical application. • While my application failed to accomplish the goals for the Asteroid Hunter testing application, it did succeed in testing the overall validity of the probability matching algorithm. • •

Source  Xie Junfeng, JiangWanshou, and Gong Jianya. A

new star identification algorithm based on matching probability. In Geoscience and Remote Sensing Symposium, 2008. IGARSS 2008. IEEE International, volume 3, pages III –1166–III – 1169, July 2008.