Research On LAMOST Fiber Position Detection Based On Front Illuminated Condition

Accurate position extraction of 4000 fiber positioners on a focal plane with a diameter of 1.75 m is a key technical challenge in the research of LAMOST closed-loop positioning system.In the closed-loop positioning system,the accurate control of fiber positioners relies on the high accuracy measurement of the fiber position by the photogrammetry system.Therefore,the accuracy of fiber position detection in the photogrammetry system is the key to determine the achievement of the closed-loop positioning system.This paper focused on the fiber position detection based on the front illumination conditions.For the task of detecting tiny fibers position under long-distance and large scale conditions,this paper tried to divide the whole detection based on frontillumination conditions into three modules in the order of detection and identification for research implementation.(1)Fiber positioners detection module:In this module,we analyzed the problem of low accuracy of Faster RCNN networks in small target detection tasks in the context of large scale images with many small targets,and proposed a pioneering small target detection method based on the "image segmentation",which solved the problem of small target miss detection in the original Faster RCNN networks.(2)Fiber position extraction module:In the study of fiber position extraction problem,this paper proposed a set of new extraction method based on improved least squares method starting from the principle of least squares fitting,which solved the problem of poor detection accuracy of the original least squares method in some special cases and improved the robustness of the algorithm.(3)Fiber positioners matching module:this paper investigated the problem of low matching efficiency of the original minimum matching algorithm under the front illumination condition,and optimized the original matching algorithm,and experimentally proved that the improved matching algorithm operated at an average of 90%higher efficiency.In order to verify the practical effect of our detection scheme,a telecentric measurement platform with low distortion and low aberration was built under laboratory conditions,and three double-head fiber positioners with standard distances were taken as the experimental detection benchmark based on this platform,and the fiber distance calculated by our detection scheme was compared with the true value measured by universal tool microscope,and the error value was used as the data of the detection accuracy of this scheme.The experimental results showed that the average detection accuracy was 9.2 microns,which met the expected requirements.In addition,in order to verify the reliability of the photogrammetry system equipped with this solution,a small focal plane pilot platform was built in accordance with the LAMOST working environment,and the fiber positioners stationary at zero position was photographed by this photogrammetry system for a long time,for a total period of one hour,and the obtained data were analyzed and the identification results of all the above mentioned modules were saved.The results showed that in this experiment,the average precision of the fiber positioners detection module was 96.5%,the recall was 100%,the standard deviation of the fiber position stability data of the fiber position extraction module was 16.7 microns,and the matching success rate of the fiber positioners matching module are 100%,all of which have achieved good results and verified the reliability of the photogrammetry system based on the scheme of this paper in real environment.