字符识别中两种改进的模板匹配算法

模板匹配算法是字符识别中最常用的一种算法。本文首先介绍了字符识别与模板匹配算法的基本原理,然后在传统模板匹配方法的基础上提出了两种改进的模板匹配算法。一是多模板匹配法,该算法可以对模糊不清、有倾斜的字符进行了有效识别;二是字符区域分割匹配法,该算法使相似度较大的字符可以得到准确的识别,并且大大减少了计算量,加快了识别速度。本文使基于模板匹配的字符识别达到了更有效,更准确的效果。     

一种基于BP神经网络的车牌字符识别算法

车牌识别系统是智能交通中的一个重要分支,本文针对传统的模板匹配算法存在识别准确率低的问题,提出了一种基于神经网络的车牌字符识别算法。该方法对分割、归一化后的字符进行特征提取获取其特征向量,把这个特征向量送到BP网络中进行训练,可以得到训练好的权值,以此权值对车牌字符进行识别。实验表明,本算法对车牌图像的识别率达90%以上。     

基于改进的LM-BP神经网络的车牌字符识别研究

车牌识别是智能交通系统中一个重要的环节,它可以应用到很多领域,如高速公路自动收费、交通监控系统、停车场管理等.提出一种改进的LM-BP神经网络车牌字符识别方法,该方法根据国内现行车牌编制的特点,结合LM算法改进传统BP神经网络,并增加σ参数修正LM-BP算法,避免传统BP神经网络收敛速度缓慢并容易陷入局部极小值的缺点,进行了大量实验,达到了预期的识别效果和收敛速度.     

基于改进的BP神经网络的车牌字符识别方法

借助于计算机控制技术中PID增量式控制算法中“加速对给定值变化的响应”思想来改进传统的BP神经网络算法。经过网络训练之后,应用其进行车牌字符识别。实验表明改进后的神经网络收敛速度和精度方面都有所提高,能提高字符识别的速度与精度。     

一种改进Transformer的仪表字符识别算法

针对传统工业仪表在智能化改造过程中,非智能仪表不能输出数字信号的问题,以数显式流量计仪表字符为研究对象,以提高仪表字符识别准确度和识别速度为目的,采用一种改进Transformer的仪表字符识别算法。该算法添加修改的EfficientNetV2s模型作为特征提取层;在Transformer模型的多头注意力中使用多头加权融合机制;并增加相对位置编码。实验结果表明,改进得到的基准模型识别精度提高到99.6%,识别速度达到107.3 ms,达到了预期的识别效果。     

复杂背景图像的字符识别算法研究

为了解决字符识别过程中的局部曝光、印刷字符的断裂以及变形和自然环境下的背景污染等问题, 提出了一种分块处理与卷积神经网络(CNN)相结合的字符图像识别算法. 首先利用OpenCV机器视觉库, 结合分块处理、伽马运算、参数调整等方法对产品零件表面印刷字符进行预处理, 初步解决图像局部曝光和字符断裂问题; 其次为了获得单个字符图像, 利用数学形态学算法对局部曝光处理后的二值化图像进行分步分割, 进而去掉字符间的无用信息; 最后利用Keras模块为字符识别提供的API搭建CNN模型, 经过对100多张字符的识别训练, 准确率高达96.9%, 为某汽车零部件自动化生产中的字符识别提供了可靠的依据.     

基于多层激励函数量子神经网络的字符识别算法

针对不同样本之间存在交叉数据的模式识别问题,将多层激励函数的量子神经网络引入模式识别之中,提出一种基于量子神经网络的模式识别算法。量子神经网络是将神经元与模糊理论相结合的模糊神经系统,由于自身固有的模糊性,它能将决策的不确定性数据合理地分配到各模式中,从而减少模式识别的不确定度,提高模式识别的准确性。本文以英文字母为例,应用量子神经网络模型进行字符识别,通过比较发现量子神经网络除了可以克服BP网络的诸多缺点外,对具有不确定性、两类模式之间存在交叉数据的模式识别问题,有极好的分类效果。仿真结果证明该方法的正确性和有效性。     

基于BP神经网络的字符识别系统

基于BP神经网络设计了一个字符识别系统.首先,对字符图像进行二值化处理,构造输入向量矩阵.其次,通过选取初始权值、隐层节点数和权值学习算法,创建BP神经网络,对样本数据进行训练,之后对加有噪声的样本再次进行训练,以提高网络的鲁棒性.最后进行仿真测试并制作图形用户界面GUI来模拟与演示该系统.仿真结果显示,该BP网络对噪声系数小于0.8的字符识别率可达95%,且网络训练时间可接受.     

基于形状上下文的字符识别算法的改进

字符识别技术在近些年来具有非常广泛的应用,同时也提出了很多识别方法。其中,形状上下文算法是一种识别率非常高的算法之一。该算法包含大量的字符信息;因此,识别速度并不快。为了提高识别速度,在形状上下文理论的基础上,结合了序贯相似性检测的思想,在非匹配的字符处缩短时间,提高整体效率。     

快速分类神经网络在字符识别中的应用

本文以海明神经网络与自适应谐振理论(ART)模型学习算法为基础,从理论上分析了海明网络学习算法的缺陷,利用ART网络的思想,提出了一种快速分类的神经元网络的算法,命名为Improved Hamming算法(简称Im-H算法)。此算法主要优点在于阈值更新及引入了经验迭代次数。将此算法用于字符模式识别,大量的计算机实验结果表明了Im-H网络学习算法的有效性、快速性。     

基于特征重构的广义回归神经网络故障识别方法应用研究

近年来,流程工业事故频发,这使得加强生产过程的安全保障迫在眉睫,对故障识别的准确性提出了更高的要求。本文提出了一种基于特征重构的广义回归神经网络故障识别方法。首先,引入"字典表"的功能构建"故障字典表":其次,采用核主元分析方法对"故障字典表"进行主元提取,实现数据降维以及降低计算复杂度;第三,"故障字典表"索引定位,通过数据样本与"故障字典表"的比对,对数据样本进行特征重构;最后,运用广义回归神经网络算法对数据样本进行学习训练,用以计算系统输出变量进行故障识别。通过对TE(Tennessee Eastman Process)过程进行故障识别仿真实验,结果表明,该方法对非线性时序系统具有较高的故障识别能力,为复杂过程工业大型系统的故障识别提供了新的思路和方法。     

Automated design of modular arithmetic devices: Might CAD replace an engineer

The results of an experimental comparison of patented modular arithmetic circuits with logic circuits synthesized by means of automated design are presented. It is shown that the patented circuits have less constructive complexity.     

Some experiments in the evaluation of a character recognition scanner

For character recognition machines in which the design of the logic is time consuming and expensive, a method for evaluating the character recognition scanner prior to designing the recognition logic is desirable. A method for evaluating a character recognition scanner by evaluating the patterns it produces is described. The method, for a given set of parameters, assigns to each pattern a figure of merit that measures the difficulty of recognizing the pattern. A number of experiments show that the evaluation method produced results comparable to results produced by a recognition logic designed by engineers. On a sample of 81,200 patterns, the evaluation method showed in this experiment that 0.20% of the patterns would be difficult to recognize, while for the same patterns, logic designed by engineers failed to recognize or misrecognized 0.24% of the patterns.     

一种神经网络硬件实现的可重构设计

以BP网络为例,提出了一种可重构神经网络硬件实现方法。通过可重构体系结构、可重构部件的设计,可以灵活地实现不同规模、传递函数及学习方法的神经网络,从而搭建起神经网络快速硬件实现的平台。经过对一个模式识别问题的实现和测试,证明了这种设计方法的可行性。     

Heuristic principles for the design of artificial neural networks

Artificial neural networks were used to support applications across a variety of business and scientific disciplines during the past years. Artificial neural network applications are frequently viewed as black boxes which mystically determine complex patterns in data. Contrary to this popular view, neural network designers typically perform extensive knowledge engineering and incorporate a significant amount of domain knowledge into artificial neural networks. This paper details heuristics that utilize domain knowledge to produce an artificial neural network with optimal output performance. The effect of using the heuristics on neural network performance is illustrated by examining several applied artificial neural network systems. Identification of an optimal performance artificial neural network requires that a full factorial design with respect to the quantity of input nodes, hidden nodes, hidden layers, and learning algorithm be performed. The heuristic methods discussed in this paper produce optimal or near-optimal performance artificial neural networks using only a fraction of the time needed for a full factorial design.     

Further results on controllability of recurrent neural networks

This paper studies controllability properties of recurrent neural networks. The new contributions are: (1) an extension of a previous result to a slightly different model, (2) a formulation and proof of a necessary and sufficient condition, and (3) an analysis of a low-dimensional case for which the hypotheses made in previous work do not apply.     

The design of the Boost interval arithmetic library

We present the design of the Boost interval arithmetic library, a C++$++$ library designed to handle mathematical intervals efficiently and in a generic way. Interval computations are an essential tool for reliable computing. Increasingly a number of mathematical proofs have relied on global optimization problems solved using branch-and-bound algorithms with interval computations; it is therefore extremely important to have a mathematically correct implementation of interval arithmetic. Various implementations exist with diverse semantics. Our design is unique in that it uses policies to specify three independent variable behaviors: rounding, checking, and comparisons. As a result, with the proper policies, our interval library is able to emulate almost any of the specialized libraries available for interval arithmetic, without any loss of performance nor sacrificing the ease of use. This library is openly available at www.boost.org.