| Large Sky Area Mulit-object Fiber Spectroscopic Telescope-LAMOST, with a 1.75m-diameter focal plate on which 4000 optical fibers are arranged, is one of major scientific projects in China. During the surveying process of LAMOST, the optical imaging system makes the astrometric objects be imaged in the focal plane, and the optical fiber positioning system controls the 4000 fibers to be aligned with these objects and obtain their spectrum. If the optical fiber positioning system works in open-loop control mode, it's difficult to ensure the positioning accuracy because of the errors caused by control, machining, installation and changes of the work environments. In order to correct these errors and make the LAMOST run accurately, it is necessary to detect the position of optical fibers in the focal plane accurately.The detection of optical fibers'position in LAMOST has the following requirements:large field, multi-discrete targets, high accuracy, rapid and non-contact, including almost all high testing standards. In the conventional position detection methods, the close-range photogrammetry has the most possibility of meeting all these requirements, so the application of the close-range photogrammetry is studied detailedly in this paper.Photogrammetry has two optional imaging devices:linear CCD and area CCD. For their advantages and disadvantages, multi-linear CCD scanning detection method and area CCD camera district detection method is proposed in this paper, and their measuring principles and processes are discussed in detail.This paper's main content includes:1. The experimental detection system, which is consisted of three linear CCD, is designed and developed. The accuracy of detecting the static optical fiber's coordinates is±0.9μm (2σ) which shows the system has a good stability.2. The position detection of dynamic optical fiber by the linear CCD system is achieved. After attitude calibration of the linear CCD, conversion splicing and optimization of parameters, the experiment demonstrates that the error of detecting the distance between the optical fibers is only±5μm (2σ) which is a high accuracy. However, the high detection accuracy requires complex calibration and optimization with a wide range of high-density reference optical fibers.3. An error analysis model of the spot location algorithm (light centroid method) is established, and the best detection condition is suggestted combined with the corresponding experiments. Under this condition, the accuracy of light centroid method is 0.04 pixels which could be used to judge the size of detection sub-region.4. An improved bundle adjustment self-calibration method is proposed to realize the on-line high-precision calibration of area CCD camera. The results of experiment indicate that this calibration method needs only a few control points while the traditional calibration methods need much more control points to get the same accuracy.The study of this paper on linear CCD method shows that it has high theoretical detection accuracy. However, it will need a wide range of reference optical fibers and high requirements on the performance and precision of hardware if the linear CCD method is applied on LAMOST in practice, and it's difficult to realize with current technology.Some key problems of area CCD method are researched and solved in this paper, so the area CCD camera district detection method can be smoothly transplanted to the locale of LAMOST from the laboratory.
|
PDF Full Text Request