Curve envelope fitting method based on VGG16 network

Abstract

The invention discloses a curve envelope fitting method based on a VGG16 network, and the method comprises the following steps: training a neural network by using a data set sample which is provided with a label and is acquired and established by a CCD, and applying the neural network algorithm to an acquired data set to verify the accuracy and calculate the microscopic morphological characteristics of the surface of an optical fiber; creating a read data set through a tenserflow framework, recording a gray value change sequence of the group of images at a certain pixel point (a, b) as X(a, b)(t), supplementing the sequence X(a, b)(t) into a one-dimensional sequence X2(a, b)(t) with the size of 224 * 224 by adopting a cubic Hermite interpolation method, and then converting the sequence X2(a, b)(t) into a two-dimensional image matrix X2(a, b)(m, n); processing and outputting the predicted actual height of the pixel point through a specially designed convolutional neural network, and comparing the actual height of the pixel point with a sample label to enable an error to be within a set threshold range. The application of the neural network enables the algorithm to have better self-learning, self-organizing and fault-tolerant capabilities and excellent nonlinear approximation capability, can improve the accuracy and fault-tolerant capability of the envelope algorithm, and has certain reference significance.

Description

technical field [0001] The invention relates to the field of deep learning data processing, in particular to an envelope fitting algorithm of a surface characteristic curve of an optical fiber ferrule based on a deep learning tool. Background technique [0002] A neural network algorithm is a class of algorithms that learn from data based on a neural network, and it still falls under the category of machine learning. With the improvement of computing power and the advent of the era of big data, highly parallelized GPU and massive data make it possible to train large-scale neural networks. In the field of image classification, the use of convolutional neural networks in deep learning is many. Compared with the traditional image classification method, it no longer needs to manually describe and extract the features of the target image, but learns features from the training samples autonomously through the neural network, and these features are closely related to the classifie...

AI Technical Summary

Problems solved by technology

At present, the commonly used peak location methods include extreme value method, center of gravity method, Fourier transform method, wavelet transform method, Hilbert transform method, space frequency domain algorithm, white light phase shift method, but the positioning accuracy and accuracy of these methods are relatively low. Difference

Images