人民币冠字号码高速识别实时模板生成算法

在回顾了目前人民币冠字号码识别研究文献的基础上,提出了一种在高速走钞环境下用于识别冠字号码的实时模板生成算法.首先在装有CIS的验钞机上采集到足够完备的人民币图像,然后抠出冠字号码区域,对其进行倾斜校正、二值化处理、形态学处理、字符分割及高度归一化后,对每种字符位图生成对应的模板集合,再根据差异最大化的原理从模板集合中挑出5个形成最终的模板精简集合.最后,给出了应用此模板进行实时识别的效果数据及分析.     

基于多结构特征的纸币字符识别算法

针对现有人民币冠字号字符识别算法对严重污染磨损的人民币冠字号字符识别率低的难题,提出了一种将模板匹配法与多种字符结构特征(字符轮廓、区域密度)相结合的人民币冠字号字符识别算法.通过定位分割、分类识别、验证纠错3个步骤完成字符的识别,实验结果表明,该算法具有较高的识别率和较强的鲁棒性.     

基于图像理解的LPR系统研究

设计了一种基于图像理解的智能车牌识别系统,利用水平和垂直投影进行车牌区域定位。通过基于垂直投影和模板匹配的字符分割方法有效地解决了背景变化、车牌尺寸变化、字迹模糊等问题,在对120幅472×332大小的车辆图像采用基于模板匹配的ORC识别方法,在Hausdorff距离分类器上进行了相关匹配测试实验。识别率达96.7%,识别时间稳定在0.3~1.2 s之间。     

基于LabVIEW的人民币冠字号识别系统研究

系统的工作方式主要是对人民币采集图像,然后进行包括图像灰值化、滤波、倾斜校正及对冠字号目标区域提取的图像预处理,随后利用LabVIEW中Vision Assistant工具里的OCR功能对冠字号中出现的字符进行训练并建立模板数据库,与提取出的冠字号区域进行匹配从而识别出字符并转换成文本。系统用户操作界面简洁,具有较高的识别准确率及速度。     

基于模板的低信噪比前视红外建筑物识别技术

针对低信噪比复杂地物背景下的前视红外地面建筑物目标自动识别问题,提出了一种基于目标模板的自动目标识别方法。首先,根据制备模板的尺度和形状信息对实时图像进行形态学背景抑制增强和目标重构处理;然后,对重构后的图像进行垂直和水平边缘线条提取和轮廓匹配,融合垂直和水平线条匹配结果得到最终的相关峰和潜在目标区域;接着,根据目标模板对潜在目标区域进行区域灰度对比度度量;最后,融合轮廓匹配和区域灰度对比度度量结果,得到最终相关峰,实现目标的识别。大量数据测试表明,该方法的正确识别率在92%以上,定位误差小于2个像素,速度快、检测性能好、适应性较强、易于硬件实现。      

变形网格及其在图像识别中的应用

网格特征是图像识别中一类重要特征,而变形模板对于复杂的图像识别问题如字符识别、数字识别、图标识别等表现出很好的性能,但变形模板很费时。该文针对网格特征提出了变形网格,并分析了变形模板与变形网格之间的近似等价性。所提方法对网格进行变形而不是对图像变形,因而速度比变形模板快得多,而且性能相差不大。把这种方法分别应用于图标识别和脱机手写汉字识别。图标识别实验中变形模板使识别率提高了7.5%,而变形网格使识别率提高了7.3%、手写汉字识别实验中变形模板使识别率提高了6.1%,而变形网格使识别率提高了5.8%。考虑到变形网格比变形模板快得多,所以这种方法是有优势的。     

基于数学形态学的扫描地图点状符号识别方法

提出了一种基于数学形态学的扫描地图符号双向模板匹配识别方法.首先通过对扫描地图进行几何校正、背景消除、对比度增强、二值化和面元去除等预处理,然后采用基于数学形态学的双向模板匹配法对预处理扫描地图符号进行识别.试验结果表明该方法不仅能良好地去除扫描地图上噪声的影响,而且能有效、可靠地识别扫描地图符号,平均正确识别率达90%以上.     

人脸画像识别研究

利用几何特征法、特征脸法和动态模板匹配法来进行画像特征提取和识别,试验表明根据文中设计的模板系统,动态模板匹配法具有较好的识别效果,相比识别率可达到82.2%.     

基于计算机视觉的集装箱号码识别系统

对基于计算机视觉的集装箱号码识别系统进行研究.针对光照不均带来的系统识别率急剧下降这一问题,提出了基于小波变换的集装箱号码定位方法.该方法使用小波变换的垂直边缘方法来过滤光照不均的干扰;同时在字符分割处理中研究了差分二值算法,根据字符边缘特性与光照影响的密集度的关系来消除光照影响.实验结果表明,该方法更有效,拓展了集装箱识别的研究与应用.     

基于Android平台的身份证号码识别

日常生活中很多场合都需要对身份证进行检查,并需要人工输入身份证号码。而身份证号码位数偏多,人工输入难免会出错而且速度慢。本文针对此问题提出了一种有效的基于Android平台的身份证号码识别算法。该算法将基于Android操作系统的智能手机的视频流捕捉功能而获得的图片作为输入图像,经过图像预处理、身份证号码区域定位、字符分割等前期步骤,然后进行字符识别。除此以外,考虑了由人工操作的随意性而带来的特殊情况,并对此作出相应的处理,提出了进一步改进的,更有效的识别算法。实验结果表明,本文的整个身份证号码识别的处理流程简单有效且识别率高,获得了满意的效果。     

一种基于交点特征的印刷体数字识别方法

为了进一步提高印刷体的数字识别准确率,提出了一种基于交点特征和径向基函数神经网络的数字识别方法。首先利用交点特征对数字进行特征提取,即提取某一数字的划水平线得到的交点数作为水平特征分量,提取划垂直线得到的交点数作为垂直特征分量,将水平特征向量与垂直特征向量组合成数字的交点特征向量;然后利用径向基函数神经网络学习不同模式类别中的学习样本,学习过程完成后,利用此网络对样本进行识别。实验结果表明,该数字识别方法在印刷体数字识别中正确率可达到100%,处理效果良好。     

Face recognition based on wavelet-curvelet-fractal technique

In this paper, a novel face recognition method, named as wavelet-curvelet-fractal technique, is proposed. Based on the similarities embedded in the images, we propose to utilize the wavelet-curvelet-fractal technique to extract facial features. Thus we have the wavelet’s details in diagonal, vertical, and horizontal directions, and the eight curvelet details at different angles. Then we adopt the Euclidean minimum distance classifier to recognize different faces. Extensive comparison tests on different data sets are carried out, and higher recognition rate is obtained by the proposed technique.     

一种基于SVM的改进车牌识别算法

提出了一种基于支持向量机(SVM)的改进车牌识别方法。对细化处理后的字符采用网格、水平投影与垂直投影密度的特征提取方法,保证了字符整体与局部特征,同时也使特征向量集的维数充分低。结合3种特征提取方法得到的特征向量集,采用 SVM进行车牌号码识别。对于易混淆字符,提出了根据各自的特征进行2次识别的算法,该算法有效解决了易混淆字符误识别的问题。实验结果表明,该算法鲁棒性好、抗干扰能力强、识别率达到了98.58%。     

Computer recognition of handwritten hebrew characters

Preliminary results of experiments with computer recognition of handwritten Hebrew characters are presented. The features used for the classification were the number of line crossings along some specified intersections, and the relative lengths of horizontal and vertical segments. Characters were clustered, first, according to some geometrical properties. Test characters within each cluster were then classified considering only those characters of the training set which were assigned to the same cluster. A min-Max decision rule yielded 90.9 percent of correct classification.     

基于字符密度的车牌识别方法研究

针对车牌识别系统的开发,提出一种基于二值图像的字符识别算法.在该算法中,采用了新的特征向量,字符的水平密度和垂直密度,分类器采用基于误差反响传播的人工神经网络算法(BP神经网络).采用这种方法可提高系统的字符识别率和系统的实时性,实验表明此算法非常有效.     

汉语连续语音识别中不同基元声学模型的复合

该文研究由不同声学基元训练的声学模型的复合。在汉语连续语音识别中,流行的基元包括上下文相关的声韵母基元和音素基元。实验发现,有些汉语音节在声韵母模型下有更高的识别率,有些音节在音素模型下有更高的识别率。该文提出一种复合这两种声学模型的方法,一方面在识别过程中同时使用两种模型,另一方面在识别过程中避开造成低识别率的模型。实验表明,采用本文的方法后,音节错误率比音素模型和声韵母模型分别下降了9.60%和6.10%。     

Vertical sectorization in self-organizing LTE-advanced networks

Cellular radio networks continuously evolve to respond the exponential growth in data traffic volume in mobile communications. Active antenna technology contributes to this evolution by introducing vertical sectorization, which splits the horizontal sector into two subsectors with respect to the elevation plane and doubles the number of cells that can be deployed. However, in order to guarantee a reliable and near-optimal operation of vertical sectorization in a system that applies universal frequency reuse, such as LTE-advanced, co-channel interference mitigation is essentially needed. In this article we propose a decentralized self-optimization method that can be used to mitigate the undesirable inter-cell interference by self-tuning the electrical antenna downtilt toward the optimal antenna elevation angle. The performance evaluations for the proposed self-optimization method are carried out for both coordinated and uncoordinated subsector transmission scenarios within the LTE-advanced framework using a dynamic LTE-advanced compliant system level SON simulator. Based on the extensive performance evaluations carried out for a realistic urban scenario, it is found that self-optimization improves the vertical sectorization performance 25 % in terms of virtual load. Furthermore, the performance gain reaches up to 30 % when vertical vectorization is provided with dynamic point selection and muting feature. Therefore, the article concludes that in LTE-advanced networks vertical sectorization can largely benefit from the antenna self-optimization and outperform the traditional horizontal sectorization approach with low algorithmic complexity.     

Resolution depths for some transmitter-receiver configurations

Equivalent dipole polarizability matrices and equivalent dipole location are a convenient way to interpret magnetic field data due to currents induced in isolated conductive objects. The uncertainties in polarizability estimates and in the equivalent dipole location provide a quantitative measure of the performance of different configurations of transmitters and receivers. In another paper, we estimate these uncertainties using a linearized inversion. For many systems, consisting of one or more rectangular loop transmitters and a number of dipole receivers, sited on a horizontal grid, equivalent dipole depth is determined to 10% accuracy to depths approximately 20% deeper than the depths at which polarizability matrix elements can be determined to the same precision. Systems that have a lower product of rms polarizability uncertainty and square root of their number of transmitter-receiver pairs are considered more effective for the number of transmitter-receiver pairs. Among the systems studied, a system with three orthogonal transmitter loops and a three-component receiver is the most effective, for objects shallower than 0.6 times the instrument siting grid spacing, yielding an rms polarizability uncertainty 0.04 times that of a single-transmitter single-receiver system. At intermediate depths, a system with two vertical component receivers on the diagonal of a square horizontal transmitter loop is most effective for its number of transmitter-receiver pairs, yielding an rms polarizability uncertainty 0.07 times that of a single receiver system. At depths greater than 2.5 times, the siting grid spacing a three-orthogonal loop transmitter with a single vertical component receiver is about the most effective for its number of transmitter-receiver pairs, yielding an rms polarizability uncertainty 0.08 times that of a single-transmitter system.     

Effect of the prefabricated routing track distribution on FPGAarea-efficiency

In most commercial field programmable gate arrays (FPGA's) the number of wiring tracks in each channel is the same across the entire chip. A long-standing open question for both FPGA's and channeled gate arrays is whether or not some uneven distribution of routing tracks across the chip would lead to an area benefit. For example, many circuit designers intuitively believe that most congestion occurs near the center of a chip, and hence expect that having wider routing channels near the chip center would be beneficial. In this paper, we determine the relative area-efficiency of several different routing track distributions. We first investigate FPGA's in which horizontal and vertical channels contain different numbers of tracks in order to determine if such a directional bias provides a density advantage. Second, we examine routing track distributions in which the track capacities vary from channel to channel. We compare the area efficiency of these nonuniform routing architectures to that of an FPGA with uniform channel capacities across the entire chip. The main result is that the most area-efficient global routing architecture is one with uniform (or very nearly uniform) channel capacities across the entire chip in both the horizontal and vertical directions. This paper shows why this result, which is contrary to the intuition of many FPGA architects, is true. While a uniform routing architecture is the most area-efficient, several nonuniform and directionally biased architectures are fairly area-efficient provided that appropriate choices are made for the pin positions on the logic blocks and the logic block array aspect ratio