Research On Arc End Defect And Array Accuracy Detection Of Optical Fiber Beam

Fiber bundle is made of a large number of single fiber arrays and is widely used due to its advantages of light weight,good flexibility and so on.In order to detect the accuracy and defects of Arc-end arrays of special-shaped array optical fiber bundles,this paper studies the design of detection scheme for Arc-end of optical fiber bundles,image splicing algorithm,array accuracy detection method and defect detection method.Firstly,on the basis of consulting a large number of relevant literatures at home and abroad,and combining the specific structure,detection parameters and relevant principles of the optical fiber bundle to be detected,the paper formulates a detection scheme,analyses the influence of mechanical structure,motion control and image acquisition on the detection accuracy during the detection process,and completes the design,selection and circuit layout of the hardware devices required in the detection scheme.Secondly,aiming at the problems that defect type misjudgment and optical fiber missing detection may occur in the process of optical fiber bundle arc end detection,the image splicing method is put forward.By analyzing common image splicing methods,the SURF algorithm based on feature matching is selected.In view of the excessive mismatch of SURF algorithm in the process of feature matching,an improved rough matching method of K-d tree nearest one-way feature points is proposed.The matching accuracy was increased by 6.8% compared with that before improvement.The connected-area center of gravity method is proposed to detect the accuracy of the optical fiber bundle Arc-end array,and the processing results of the optical fiber bundle Arc-end image with different graying and filtering methods are analyzed.The optical fiber centre of gravity at the optical fiber bundle Arc-end is obtained based on Otsu image segmentation and morphological repair.Then,the common defect types and existing detection methods are discussed,the data sets of common defect types are made and expanded,and an improved YOLOv5 s network model is put forward for the detection of defects at the arc end of optical fiber bundle.That is,CBAM attention mechanism is added to the original YOLOv5 s network model,and EIOU boundary frame loss function is used instead of CIOU boundary frame loss function,which enhances the ability of identifying obvious defects before improvement.The speed of convergence is accelerated.Finally,the detection experiment of the arc end defect of the optical fiber bundle is carried out by using the established detection platform,and the detection experiment of the arc end array accuracy of the optical fiber bundle is carried out based on the connected area gravity center method,and the test results are analyzed to verify the correctness of the detection method studied in this paper.