Research On Leo Target Detection For Binding Telescopes With Large FOV

In view of the importance of space target detectoion in the domain of space surveilliance,the researchers have devoted themselves to the technology of target detection.So far,the research on moving space target detection for large field of view(FOV)are still facing a lot of difficulties,which are from either the sensor itself,or the characteristics of the target and the background.Motivated by the demand for moving target detection in full-frame sensor images of earth-based binding-telescope detectors with large FOV,this paper deeply and systematically studies and explores the techniques of space target detection.The work of this paper is organized as follows.Chapter 2 firstly introduces the conception and data processing approaches of binding telescopes,and analyzes its signal-to-noise ratio(SNR)advantage,Then,the information processing flow of the space target detection system is given,such as mage pre-processing,star identification and space targe detection.At last,the chapter introduces the knowledge of celestial-body measurement that space target detectors involve,including the definition of different kinds of space reference systems and the transformation relationships between them.The knowledge lays foundation for the statements in follow-up chapters.Chapter 3 is devoted to the pre-processing of the full-frame images.First,diferent kinds of noise in optical instruments are introduced and analyzed.Then,the morphology algorithm is employed to suppress the background,hence increasing the SNR of the image.On basis of this,in respose to the smear problem of full-frame imges,we propose a smear removal algorithm based on Gaussian fitting.Processed by the algorithm,the smears are basically eliminated and the star images are preserved.The smeal removal procedure provide images with higher equality for later star extraction,identification and target detection,to eliminate large amount of false alarms.Chapter 4 mainly studies star identification techniques for the purpose of eliminating the false alarms that the stars in the FOV cause in space target detection.This is due to the fact that stars are the main background objects in the FOV as well as the most interference in space target detection.Therefore,in response to the problem of star pattern recognition in cameras with large FOV,a star identification algorithm based on region growing is proposed.Different from traditional methods,the proposed algorithm does not adopt overall strategy to perform identification,but recognizes stars in a local area where the distortion error is limited,and extends the recognized areas gradually until the entire image.The algorithm can present high identification rate compared with other methods,while the high compution speed is maintained.Chapter 5 firstly proposed a single-frame detection method based on clustering algorithm and directional growing.The algorithm makes use of the motion characteristics of space targets and the features of gradient smears of stars in full-frame images,calculates the slopes of probable object blocks using the weighted least square method,and obtains candidate targets by clustering.The suspected targets can be acquired by searching gradient target smears in the neighborhood according the smear slope calculated with the motion features of the target.Based on the results of the single-frame detection,the PHD algorithm is exploited to perform sequential processing on the measurement vector of the target,and regards the output target situation as the multi-frame detection results of the space targets.The experimental results show that the algorithm presents good performance of detection,and the multi-frame detection procedure effectively decreases the false alarm propability.