基于神经网络的二维码识别算法

Two-Dimensional Code Recognition Algorithm Based on Neural Network

摘要: 本文主要设计一种残缺二维码的识别算法，通过设计具备记忆能力的神经网络，建立残缺二维码与电缆信息的映射关系，从而在二维码因外界原因出现残缺的时候，能够读取与之对应的电缆信息。该算法中的神经网络采用动力学的Lyapunov函数，利用每个节点的状态变化记忆源二维码的像素信息，首先，利用二维码扫描装置，将二维码的图像信息传递给二维码识别模块；然后，将二维码图像转化为二值像素矩阵，作为神经网络的输入；接下来，利用源二维码的像素矩阵训练神经网络各个神经元之间的连接权值，得到相应的权值矩阵；随后，输入残缺二维码的像素矩阵之后，各神经元节点不断调整状态值，直到各神经元状态值的变化小于阈值之后，停止调整；最后，将神经元状态值矩阵对应源二维码的电缆信息，从数据库中读取出来。

Abstract: This paper mainly designs a recognition algorithm of incomplete two-dimensional codes. By de-signing a neural network with memory ability, the mapping relationship between incomplete two-dimensional codes and cable information is established. Thus, when two-dimensional codes are incomplete due to external reasons, the corresponding cable information can be read. In this algorithm, dynamic Lyapunov function is used to memorize pixel information of two-dimensional code by the state change of each neural network node. Firstly, image information of two-dimensional code is transmitted to two-dimensional code recognition module by two-dimensional code scanning device. Then, two-dimensional code image is transformed into a binary pixel matrix as a neural network. Next, the connection weights between neurons are trained by pixel matrix of source two-dimensional code to get corresponding weight matrix. Then, after the pixel matrix of incomplete two-dimensional code is input, neuron nodes adjust the state values until change of neuron state values is less than threshold value. Finally, cable information corresponding to the source two-dimensional code is read out from database.