| The Sky Area Multi-Object Fiber Spectroscopic Telescope(LAMOST),which is also named by Guoshoujing Telescope,is the first astronomical national major scientific facility with both large field of view and large aperture designed and built in China.It has long had the highest spectrum acquisition rate in the world over 10 years since its operation in 2009.The fiber positioning technology is one of the two key technologies of LAMOST.It included 4000 robot-like fiber positioners,and these positioners drive the fibers to fine track the images of the celestial objects collected by LAMOST and transmit as much celestial light as possible through fibers to the corresponding spectrographs for spectrum survey.The fiber positioning system is operated in open-loop mode.The accuracy of each fiber positioner is greatly affected by many factors,such as the motion accuracy of stepping motors,the accuracy of gear machining,the thermal deformation of focal structure,and the atmospheric turbulence in telescope enclosure.It is difficult to satisfy the requirements that the final open-loop positioning error of 4000 optical fiber positioners should be less than 40 microns.With the further improvement of the technical requirements of the new generation spectroscopic survey telescope,the size of the optical fiber positioner is reduced to be about 10mm,and the corresponding fiber positioning error is required to be improved about 10 microns.Therefore,it is necessary to develop a highaccuracy real-time fiber detection technology for positioning thousands of optical fibers on a large-scale focal plane.At present,LAMOST is upgrading its optical fiber position detection system into a close-loop one by introducing a fast high-accuracy real-time close-loop optical fiber position detection system.Such system can realtime feedback the 4000 fiber position errors to the fiber positioners control system to realize iterative close-loop position correction until final expected accuracy is reached.It will greatly improve the operation efficiency and positioning accuracy of the LAMOST fiber positioning system.In order to further develop our close-loop fiber positioning technology and meet the accuracy requirements of optical fiber positioning system during LAMOST observation,a close-loop high-precision detection technology of optical fiber position for the next generation spectroscopic survey telescope is focused and studied as follows:1.According to the structural characteristics of the LAMOST telescope,a set of long-distance high-accuracy real-time optical fiber position detection system was designed.It will utilize six cameras installed around the LAMOST primary mirror,which is about 20m away from the focal plane and fibers,detect the 4000 optical fiber positions simultaneously,and feedback all the fiber position errors to the fiber positioners control system to realize iterative close-loop position correction until final accuracy of 40microns is reached.Because of high requirements for the fiber position detection accuracy,the key factors affecting the accuracy in the detection system are analyzed to achieve the accuracy less than 40 microns.2.The influence of lens distortion on the detection accuracy of photogrammetric system is analyzed,and a customized method to correct the influence of lens imaging error by using high-accuracy matrix grid standard target was proposed to improve the measurement accuracy of photogrammetric system.A model of high-accuracy optical fiber position detection system was established,and a series of experiments were carried out to study the influence of lens distortion in different field of view of bi-telecentric lens,and the variation conditions of distortions of large aperture bi-telecentric lens in different field of view were obtained.By using the standard target at 75mm×75mm field of view and polynomial fitting method with 5 parameters,the large aperture telecentric lens error was corrected in a series of experiments,and the detection error was reduced well within 10 microns.The corrected bi-telecentric lens detection system was applied to carry out the closed-loop detection experiment under the laboratory conditions.The results verified that the method based on above improved bi-telecentric lens detection system of optical fiber position as feedback and iterative positioning correction could achieve accuracy better than 10 microns under the laboratory at close range,and met the requirements of optical fiber positioning system.3.A set of error correction system based on fiducial fiber is established,and it solved the problem of system instability caused by atmospheric disturbance and temperature change in long-distance and large-scale conditions.In order to study the influence and correction of detection system error under long-distance conditions,a high-accuracy optical fiber position detection system was established for a small focal plane system 20m away with a diameter of about 500mm to simulate the LAMOST real optical path conditions in the laboratory.The further experiment results verified that after correction of the fiducial fiber measurement error of the detection system by using the polynomial fitting with 20 parameters,and the detection accuracy can reach 10 microns in LAMOST conditions.By further adoption of standard target or fiducial optical fiber as the measurement basic framework of the whole detection system,stable and reliable error correction parameters could be obtained.The experimental results shown that based on the measured fiducial fiber position and parameters from either the laser tracker or the large standard target and then the calculated small standard target residual error,the detection accuracy of the system was less than 10 microns.The detection system was tested on the small focal plane,the positioning accuracy of total 235 optical fiber positioners positioned based on both random points and 84 standard matrix grids could meet the original system specification of positioning accuracy less than 40 microns after three corrections.4.A high-accuracy image recognition method for optical fiber based on frontillumination was proposed to quickly identify the operation status of the optical fiber positioner and provided instant warning of the positioners’ collision.It is impossible for back-illumination detection method to decide the operation status of fiber positioner and the subsequent dynamic programming to avoid of collision.Therefore,a high-accuracy optical fiber position recognition method based on front-illumination was proposed based on the improved Hough algorithm,to accurately recognize the cylindrical ceramic face target on LAMOST fiber positioner,which hold the optical fiber through the central hole.Therefore,it can be adopted to evaluate the operation status of fiber positioner,give collision warnings,reduce the current system complexity and increase the system operation reliability.5.Combining the deep learning method with Hough image algorithm,the recognition accuracy of optical fiber end face based on front-illumination was improved,and the problem of multiple optical fiber recognition and detection was solved.In a high-resolution image,it is very difficult to test and screen out the small optical fibers for large-scale front-illumination fiber position detection application.In order to solve the influence of the metal parts reflection and other stray light to the fiber positioner during the front-illumination optical fiber position detection,deep learning algorithm and deep learning network training can help quickly and accurately detect the positions of all optical fiber ceramic faces from the complex image environment.The proposed front-illumination detection method has broken through the bottleneck of large-scale frontillumination optical fiber position detection,achieved high-accuracy laboratory detection accuracy,and further effectively promoted the development and improvement of front-illumination detection technology. |
PDF Full Text Request