基于Simhash算法的海量文本相似性检测方法研究

为了在知识文档搜索中更加精确的为用户推荐更多语义内容相似的文档。本文对基于Simhash算法的文档相似性计算技术进行深入研究,引入ICT-CLAS分词技术,将TF-IDF技术作为计算权重的主要方法,对原有的Simhash算法作出改进,采用海明距离对Simhash指纹值进行相似性度量计算。最后以民机研制领域的工序数据为实验数据进行相关实验,实验结果表明:改进的方案性能得到提高,并且总体优于Shingle算法和原Simhash算法,能够实现大规模文档中相似性的精确检测。     

使用粒子滤波和差分进化法实现轮廓跟踪

退化现象可能导致基于粒子滤波(PF)的轮廓跟踪误差增大甚至失败。为解决这一问题,提出一种将PF状态向量后验概率密度估计问题转化为最优化问题,并使用差分进化法(DE)求解轮廓跟踪算法。该算法的最优化目标函数定义为轮廓模型与运动目标边缘匹配误差;函数参数为运动参数向量;PF的状态向量分布空间作为DE的搜索空间。目标函数的最优解对应着运动参数的最优估计值。在仿真图像序列的轮廓跟踪实验中,重心位置误差不大于0.5pixel,每帧图像平均耗时约80ms。     

基于贝塞尔轨迹的视觉导引AGV路径跟踪研究

为提高视觉导引自动导引车(automatic guided vehicle,AGV)路径跟踪精度,提出一种基于贝塞尔轨迹的精确路径跟踪算法。该算法首先将采集的多种路径特征图像进行预处理得到形状信息,训练SVM多层路径形状特征分类器;然后根据命令,改变AGV采集到的分支路径图像的权重,迭代计算所选择路径的若干最小内接圆;最后,利用最小二乘规则,将最小内接圆的圆心拟合成贝塞尔轨迹,实现AGV的精确路径跟踪。将该算法应用于视觉引导AGV中,并进行路径特征的在线识别和轨迹跟踪实验。结果表明:路径特征的识别准确率为99.7%以上,识别时间约为22 ms,弯道轨迹跟踪准确度为20 mm和20°;与传统方法相比,该方法显著提高路径特征识别和轨迹跟踪的准确率,更能满足工业现场需求。     

结合水平集和粒子滤波的人脸轮廓跟踪

针对视频序列中的人脸轮廓跟踪问题,提出了一种将改进水平集方法引入到粒子滤波框架中的轮廓跟踪算法.此算法使用零水平集函数对人脸轮廓进行表示,通过水平集函数的演化对人脸轮廓进行逼近.为了解决人脸遮挡问题,将形状先验加入到演化过程中来约束曲线的演化.算法在粒子滤波的框架下可以同时跟踪人脸区域的仿射运动和人脸轮廓的形变.实验结果表明,所提出的'方法可以对人脸轮廓进行精确的跟踪,并对外界的光照变化,背景干扰,人脸部分遮挡有较强的鲁棒性.     

采用邻域差值特征模板的立体匹配方法

提出了一种基于邻域差值特征模板的立体匹配方法。该方法通过计算一幅图像中间隔一定距离的两点的像素邻域差值作为特征模板,再计算另一幅图像中在同一扫描线上间隔同样距离的像素邻域的差值,将它与特征模板的偏差作为匹配标准,偏差最小的点就认为是匹配点。在同等条件下,与区域相关法相比,该方法能够将运算的速度提高3倍,且同样能够得到精确浓密的视差图。实验证明,该算法结构简单,易于实现,能够处理复杂的场景,具有良好的实验效果和实用价值。     

基于场景平面结构的增强现实方法

提出了一种基于场景平面结构的增强现实方法.首先,选定图像序列中的两帧对目标场景进行局部的三维重建;然后,选择场景中某一平面作为处理和跟踪的对象,通过交互指定该平面内矩形的四个顶点作为虚拟对象的嵌入区域并建立局部参考坐标系,从而实现虚拟对象与三维场景的迅速对齐;最后通过鲁棒的场景平面跟踪算法实现虚拟对象在整个图像序列中的真实感嵌入,完全允许嵌入区域被部分遮挡的情况.真实序列图像的实验结果及对算法的分析验证了方法的有效性和可行性.     

基于Potrace点跟踪的文创产品逆向设计方法研究

目的 为辅助设计人员进行产品创新设计,特别是针对文创产品设计领域提高设计效率和产品快速迭代的需求,利用产品逆向设计方法,对设计实物二维形态进行快速扫描,获取其外部形态轮廓数据,在此基础之上进行改良或者创新设计,对于快速生成产品创意和设计实践都提供了新的方法和思路。方法 针对逆向设计中,轮廓提取细节缺失、弧度线条生硬问题,本文提出了一种基于Potrace点跟踪的图像矢量路径优化算法用于文创产品的数字化分析及逆向设计。此方法以点跟踪作为基础,先利用位图中的顶点将图像轮廓抽象成多边形路径,然后再根据多边形转角大小利用贝塞尔曲线进行优化,最后通过交叉熵损失函数优化最终路径。结果 实验表明,算法在复杂边缘的矢量化线条过渡好于其他算法,并且算法与目前常见矢量化算法相比较降低了时间复杂度,算法运行时间较其他算法能够提升40%以上。结论 针对Potrace方法,提出了一种新的损失函数优化算法,算法在复杂边缘处过渡优于其他算法,同时算法运行时间大大降低,有利于将此算法用于文创产品逆向设计方法研究中。     

基于Gabor-统计特征与SVM的文档图像文本检测方法

目的对文档图像中的文本进行精确检测,深入研究统计特征对于文字纹理特征分类的影响。方法首先结合Gabor-统计特征获得文档图像的特征图像,再应用SCA算法提取文本样本和非文本样本,最后采用SVM实现文本检测,而统计特征的选择使用Fisher准则实现。结果依据Fisher准则,逆差距特征对于Gabor特征分类的类间离散度最大,效果最佳。结论针对不同类型的文档图像,使用Gabor-逆差距特征能够获得较好的检测效果。     

基于目标轮廓的附着物定位与剔除方法

工业现场清洁过的微型工件表面仍会有少量灰尘、发屑等附着物存在,在微型工件的视觉检测系统中会因其改变提取的目标轮廓而影响检测结果.为此,以灰尘与工件存在较小差异的任意位置和形状的一类附着物为考察对象,以区域生长提取的目标轮廓为先验背景,研究附着物定位与剔除算法.首先,荻取沾染附着物的工件图像,采用基于区域的分割算法做处理,以建立工件轮廓的先验知识;其次,从曲率角度定位附着物轮廓角点,以此剔除附着物轮廓;最后,根据先验知识自动修复断开的外轮廓.实验结果显示,加入附着物去除与修复算法后测量精度没有降低,测量结果误差6 μm以内,图像边缘的定位准确度能够给予保证.表明所研究的附着物定位与剔除方法使检测系统在允许微小附着物存在并且不影响测量精度的情况下,实现了目标轮廓的正确判别,提高了视觉检测系统的可靠性.     

一种新的红外与可见光图像融合与跟踪方法

多传感器图像融合技术在目标检测与跟踪领域中有广泛应用。提出了一种红外与可见光图像的新颖的特征级融合与运动目标跟踪方法。将目标轮廓用动态轮廓线表示,在目标跟踪过程中对于两类模式图像中的目标轮廓控制点向量,利用 B 样条形状空间模型将目标轮廓的特征级融合转换为控制点向量差的 L2 范数平方极小化。这种方法不需要图像配准,降低了融合的计算复杂度。同时,使用了自适应 Kalman 滤波技术,提高了动态轮廓线特征搜索的准确性。对比跟踪实验表明,融合后可见光图像的平均跟踪误差减小了 56.96%。     

基于直观汉字构形原理的C3-GAN字体生成优化方法

目的 为了提升生成对抗网络汉字风格迁移的图像生成质量,实现汉字智能生成在字库产业中的实际应用,提出了一种基于直观汉字构形学的条件生成对抗网络字体生成优化方法(Optimization of Conditional Fonts Generation with Chinese Character Configuration GANs,C³-GAN)。方法 建构了直观汉字构形模组(C³ Module),该模组包含了利于条件生成对抗网络进行汉字构形语义特征学习的全特征汉字字符集。C³-GAN在条件生成对抗网络模型下进行字体生成训练,降低了必要训练样本数量,实现对字体生成效果的优化。结果 使用C³-GAN生成汉字图像的清晰度更高、字形更准确。在图像相似性定量评估中,使用C³-GAN的实验组相比于其他模型,获得了更高的相似值和更小的误差值。结论 使用C³-GAN可以降低必要训练样本数量、提升汉字图像质量。在实际项目中具有一定的应用性和可操作性。     

Application of glyph-based techniques for multivariate engineering visualization

This article presents a review of glyph-based techniques for engineering visualization as well as practical application for the multivariate visualization process. Two glyph techniques, Chernoff faces and star glyphs, uncommonly used in engineering practice, are described, applied to the selected data set, run through the chosen optimization methods and user evaluated. As an example of how these techniques function, a set of data for the optimization of a heat exchanger with a microchannel coil is adopted for visualization. The results acquired by the chosen visualization techniques are related to the results of optimization carried out by the response surface method and compared with the results of user evaluation. Based on the data set from engineering research and practice, the advantages and disadvantages of these techniques for engineering visualization are identified and discussed.     

Information hiding technique with double phase encoding

We propose a technique for information hiding using double phase encoding. The proposed method uses a weighted double phase-encoded hidden image added to a host image referred to as the transmitted image. We develop an analytical presentation for the system performance using the statistical properties of double phase encoding. The peak signal-to-noise-ratio metric is used as a measure for the degradation in the quality of the host image and the recovered hidden image. We test, analytically, the distortion of the hidden image that is due to the host image and the effect of occlusion of the pixels of the transmitted image (that is, the host image containing the hidden image). Moreover, we discuss the effect of using only the real part of the transmitted image to recover the hidden image. Computer simulations are presented to test the system performance against these types of distortion. The simulations illustrate the system ability to recover the hidden image under distortions and the robustness of the hidden image against removal trials.     

Statistically optimised method for detecting adult image groups

Adult image recognition is an important technique for preventing children from accessing offensive material on the Internet. Most of the related works focus on single image recognition. However, adult images usually exist as a group and rarely stand alone. Therefore, considering the entire image group as a whole for classification should be more effective. This paper presents a new method of detecting adult image groups, which aims at achieving optimal recognition accuracy. Adult image group recognition generally includes two components: an adult image recogniser and a final decision rule for classifying the image group. We provide a theoretical analysis to clarify the correlation of the two components and use probability formulae to estimate the recognition rates for different settings of the adult image recogniser and the decision rule. Then, a set of optimal receiver-operating characteristic (ROC) curves for different image numbers is solved. To recognise an unknown image group, a desired recall rate for adult (or benign) image groups is specified and the system is set according to the parameters acquired from the optimal ROC curves. The proposed method can be dynamically adapted to the recall rates that the user expects. This advantage makes the proposed system more suitable for real applications. Our work can be viewed as an extension of single image recognition and the experimental results demonstrate that it can attain higher recognition accuracy than the earlier methods.     

Optimal reduction of noise in image processing using collaborative inpainting filtering with Pillar K-Mean clustering

Digital image processing is a mechanism for analysing and modifying the image in order to improve the quality and also to manage the unwanted involvement of noises. In image processing, noise is characterized as an unwanted disturbance which occurs while capturing the actual image thus affecting the quality of the image. Hence, noise formation is considered as a perilous issue and the reduction of noise is considered as an awkward process. Nowadays, almost in all fields of science and technology, digital image processing is increasing rapidly, so there arises the need for de-noising to cure the noised image. The main objective of this paper is to overcome the issue of noise and also to increase the quality and pixel value of the image. An advanced methodology known as collaborative filtering and Pillar K-Mean clustering is discussed in this paper to overcome the abovementioned problem. Initially, distinct pure images are taken as the dataset and three types of noises are added to the corresponding image to make it as a noised one. Hence, the unspecified noise is resolved on the basis of a hybrid combination of algorithms of collaborative filtering with the image inpainting method. Sequentially, the low-density noises, such as random noise and poison noise, are recovered by the implementation of collaborative filtering, and the high-density salt and pepper noise are recovered by the image inpainting method. Based on the GLCM (Grey Level Co-occurrence Matrix) feature, the normal image and the noised image are used for the clustering process. Then the de-noised image is evaluated to find the efficiency on the basis of few parameters such as SNR (Signal to Noise Ratio), MSE (Mean Square Error), PSNR (Peak Signal to Noise Ratio) and SSI (Structural Similarity Index). Accordingly, the evaluated images are further withstood for clustering to differentiate the noises by applying the proposed clustering methodology. Then the evaluated images are verified on the basis of a few parameters such as Silhouette Width, Davies–Bouldin Index and Dunn Index. The proposed methodology is run on the platform of Mat Lab. Finally, the proposed methodology is considered as an efficient method for settling the issue in digital image de-noising.     

A web-based digital analysis interface for image quality assessment

Image processing and transmission systems may introduce some amount of distortion or artefacts in a digital image. This fact usually leads to a visual or statistical image quality assessment (IQA) required in many applications and research studies in order to analyse a product image in terms of deteriorations as well as effects of the processing. There are numerous IQA criteria presented in the literature separately that makes it really difficult both for use in comparative analyses and for educational purposes. In this presented work, a user friendly web-based digital educational interface for full-reference or no-reference image quality assessment using MATLAB builder NE has been developed. In addition to this, developed system performs user-defined optional deteriorations (median noise, Gaussian blur, motion blur, high-pass filter and jpeg compression) on the image and assesses deterioration of the image. It is an extremely easy, fast and economical way of analysing digital images, especially designed for researchers, graduate and postgraduate students who work on digital image processing. Using this web-based tool well contributes to teaching all of the IQA methods and quality effects of systematic distortions on the image as well as establishes a scientific benchmark for researchers.     

USB接口的嵌入式图像采集与显示系统

介绍一种基于USB接口在嵌入式Linux开发环境下以三星S3C2410为系统处理核心、以CMOS图像传感器OV7620为图像采集工具、以USB图像处理器OV511为图像处理工具来实现图像采集和显示的方法.     

Recognition and Classification of Pomegranate Leaves Diseases by Image Processing and Machine Learning Techniques

Disease recognition in plants is one of the essential problems in agricultural image processing. This article focuses on designing a framework that can recognize and classify diseases on pomegranate plants exactly. The framework utilizes image processing techniques such as image acquisition, image resizing, image enhancement, image segmentation, ROI extraction (region of interest), and feature extraction. An image dataset related to pomegranate leaf disease is utilized to implement the framework, divided into a training set and a test set. In the implementation process, techniques such as image enhancement and image segmentation are primarily used for identifying ROI and features. An image classification will then be implemented by combining a supervised learning model with a support vector machine. The proposed framework is developed based on MATLAB with a graphical user interface. According to the experimental results, the proposed framework can achieve 98.39% accuracy for classifying diseased and healthy leaves. Moreover, the framework can achieve an accuracy of 98.07% for classifying diseases on pomegranate leaves.     

论形象思维在漫画艺术创作中的作用

1.形象思维陈学迅在《论形象思维的定义及其基本思维形式》中提出:“形象思维是创造形象,是对艺术美的思维,其核心问题就是要把形象创造出来.”[1]形象思维就是创造形象的思维,形象思维的结果就是创造一个新形象.     

A study of the feature‐based digital image stabilization system

Abstract

This article presents a digital image stabilization scheme that uses image processing techniques to compensate for undesirable image jitter due to vehicle or platform vibration to obtain a stabilized image display. The developed digital image stabilization system is based on the image combination approach combined with advanced image selection, feature detection, feature matching and pre‐rotation processing to produce crisp images. Using the Sobel edge detector, we calculate the magnitude of edge response as the judgment criteria for image transformation. To demonstrate the validity of the proposed technique, experimental results are given for some synthetic images.