基于FREAK算符及双向最近汉明距的非理想虹膜识别算法

针对低约束条件的非理想虹膜识别性能的显著下降,以及尺度不变特征变换(SIFT)和加速稳健特征(SURF)方法实时性不高的问题,提出FREAK算符配合双向最近汉明距的识别算法。通过高斯核构建多尺度特征点检测算子提取稳健的特征点集合,并引入FREAK算符改进SIFT算符,以提高特征点的表征和匹配速度;利用双向最近汉明距的匹配策略增强特征匹配对的稳定性,以降低非同源虹膜的匹配点数。实验结果表明,识别等错误率和正确识别率均有改善且单次验证时间均在0.3 s左右。该方法能够有效应对非理想虹膜类内纹理质量的变化,与SIFT与SURF算法相比具有更好的实时性。

A Novel Method for Forecasting Freak Wave Occurrence in A Short Crested Coastal Sea

In order to forecast the distribution of crest amplitudes and the occurrence of freak waves in a short crested coastal sea,a novel transformed linear simulation method is initially proposed in this paper.A Hermite transformation model expressed as a monotonic cubic polynomial serves as the foundation for the novel simulation technique.The wave crest amplitude exceedance probabilities of two sea states-one with a directional wave spectrum based on the measured wave elevation data at the Yura coast and the other with a typical directional JONSWAP wave spectrum-have been predicted using the novel simulation method that has been proposed.The likelihood that a particular critical wave crest amplitude will be exceeded is directly correlated with the probability that freak waves will occur.It is shown that the novel simulation approach suggested can provide predictions that are more precise than those obtained from the Rayleigh crest amplitude distribution model,the Jahns and Wheeler crest amplitude distribution model,or the conventional linear simulation method.This study also demonstrated that the nonlinear simulation method is less effective than the novel simulation method in terms of efficiency.

一种SIFT-FREAK图像匹配算法

图像匹配算法是计算机视觉应用研究的基础。为了解决传统尺度不变特征变换算法(Scale-invariant Feature Transform, SIFT)运行效率低、存在误匹配点对和精匹配点对稀少等问题,本文将其与快速视网膜关键点(Fast Retina Keypoint, FREAK)算法和PROSAC算法相结合,提出了一种SIFT-FREAK图像匹配算法。首先在特征点检测阶段,用SIFT算法提取具有尺度不变性的特征点,然后利用FREAK算法构建二进制描述子,特征匹配时,先采用汉明距离(Hamming Distance)进行初始匹配点对,再用双向匹配完成粗匹配,最后用PROSAC算法进行精匹配。实验结果表明,本文SIFT-FREAK算法比SIFT算法和FREAK算法(以加速分割检测特征(Feature from Accelerated Segment Test, FAST)算法提取特征点)在准确率、运行效率和精匹配点数三大方面都有着一定的优势。

Experimental Study on the Wavelengths of Two-Dimensional and Three-Dimensional Freak Waves

Freak waves are commonly characterized by strong-nonlinearity, and the wave steepness, which is calculated from the wavelength, is a measure of the degree of the wave nonlinearity. Moreover, the wavelength can describe the locally spatial characteristics of freak waves. Generally, the wavelengths of freak waves are estimated from the dispersion relations of Stokes waves. This paper concerns whether this approach enables a consistent estimate of the wavelength of freak waves. The two-(unidirectional, long-crested) and three-dimensional(multidirectional, shortcrested) freak waves are simulated experimentally through the dispersive and directional focusing of component waves, and the wavelengths obtained from the surface elevations measured by the wave gauge array are compared with the results from the linear, 3rd-order and 5th-order Stokes wave theories. The comparison results suggest that the 3rd-order theory estimates the wavelengths of freak waves with higher accuracy than the linear and 5th-order theories. Furthermore, the results allow insights into the dominant factors. It is particularly noteworthy that the accuracy is likely to depend on the wave period, and that the wavelengths of longer period freak waves are overestimated but the wavelengths are underestimated for shorter period ones. In order to decrease the deviation, a modified formulation is presented to predict the wavelengths of two-and three-dimensional freak waves more accurately than the 3rd-order dispersion relation, by regression analysis. The normalized differences between the predicted and experimental results are over 50% smaller for the modified model suggested in this study compared with the 3rd-order dispersion relation.

基于FREAK特征的多视角影像匹配算法研究

以特征点为基础的影像匹配算法具有稳定性好、效率高等优点。特征点提取易受光照、尺度以及旋转等因素的影响,当同名特征点的准确率较低时,将直接影响后续的影像拼接、目标识别等工作的正常进行。针对不同旋转角度影响影像匹配效果的问题,提出了基于FREAK特征的多视角影像匹配算法,该算法首先基于SIFT实现特征点的检测,并利用FREAK描述子对提取的特征点进行描述;其次,采用比值提纯法对特征点进行粗配准;最后,采用RANSAC算法对粗配准结果作进一步筛选,得到两幅影像的精确匹配结果。为了验证该算法的有效性,基于2组不同旋转角度的影像数据,分别对SURF算法、SURF-FREAK算法和本文提出的算法进行了对比实验,结果表明:本文提出的算法具有较高的准确率,同名点精度相对更高,对不同旋转角度的影像匹配具有较好的稳定性。

基于改进FREAK的增强现实实时注册算法

为提高增强现实实时注册的鲁棒性,针对FREAK特征描述符的特点,提出一种IFREAK特征描述符.该描述符是基于FREAK描述符的取样模型基础上考虑了空间结构的描述符.IFREAK描述符中,每一感受野对存储多个位,位数取决于参数.相比于FREAK描述符的存储方式,IFREAK描述符提高了特征点匹配的精度.在IFREAK描述符基础上本文提出一种改进的增强现实实时注册算法.实验表明,本文提出的IFREAK特征描述符的匹配度在大多数场景下优于FREAK描述符及其他描述符.本文提出的基于IFREAK的增强现实实时注册算法具有更强的鲁棒性.

关于一种危害船舶安全的波浪—Freak波的探讨

波浪是研究船舶稳性、可靠性、安全性和设计船舶的重要环境条件之一。对于它的认识和了解还远没有完成 ,尤其是对一些特殊的波浪。Freak波就是其中之一。Freak波是一种仅发生在某些海域 ,但危害性巨大的波浪 ,近年来逐步为人们所重视 ,但国内对它的研究还比较少。本文在分析国外一些研究成果的基础上 ,对比总结出Freak波的特征 ,并根据以往的观测资料总结了Freak波对船舶的危害。本文重点是引入完全非线性控制方程Schr dinger方程 ,对Freak波进行模拟 ,为定量地研究它对船舶的危害打下理论基础。

基于改进的FREAK算法的图像特征点匹配

针对采集图像时存在尺度、光线强度以及图像间旋转的差异等现象而导致图像匹配不理想的问题,提出了改进的FREAK算法(SURF-FREAK):将SURF算法和FREAK算法结合进行图像匹配。首先,利用Hessian矩阵确定候选点,进行非极大值抑制,提取的特征点具有尺度不变信息;然后用FREAK描述子对特征点进行描述,为特征点分配方向;最后利用最近邻算法进行图像配准。改进算法经实验与传统SIFT、SURF、以及FREAK算法比较分析后得知,在图像特征点匹配准确度上有一定的提高,并且对图像的尺度差异、光照差异以及旋转差异具有良好的鲁棒性。

非线性freak波分析及模拟理论探讨

概述了国内外学者对freak波的认识,并从各种观测资料分析总结了freak波的特征。引入完全非线性控制方程Schr dinger方程的1+1模式,对freak波进行模拟,为在实验室造波或计算研究提供一种方法。

基于改进的FAST和FREAK的图像匹配算法

针对传统图像匹配算法匹配时间较长、误匹配率较高的问题,提出一种改进的FAST和FREAK的图像匹配算法.该算法首先在圆形邻域上不断改变像素点个数,并与其他FAST像素模板进行对比,从而建立FAST-9特征点提取方法;然后计算其FREAK局部不变特征描述符,生成特征向量;最后通过RANSAC一致性筛选剔除误匹配点.实验结果表明,本文算法与SIFT、BRIEF算法比较,图像匹配时间缩短且图像匹配精度有一定的提高,并且对图像的旋转差异、尺度差异和光照差异都具有较好的鲁棒性.

基于改进的FREAK算法的机器人抓取系统

为了解决工业机器人对带有复杂纹理的平面型工件的识别与定位问题,将机器视觉技术与工业机器人相结合。利用改进的FREAK算法对目标工件进行快速识别,并根据工件的形状和单应矩阵计算出工件的形心坐标和旋转角度,再结合双目视觉系统,获取机器人抓取点的三维位姿信息,引导SCARA机器人抓取目标工件。实验结果表明:在复杂的工业环境下,机器人抓取系统能够准确识别出目标工件,具有良好的鲁棒性,同时也满足工业生产的实时性要求。

Freak wave simulation based on nonlinear model and the research on the time-frequency energy spectrum of simulation results

VOF （volume of fluid） method has been used to make the numerical simulation of freak wave come true. The comparisons between the numerical results and linear theoretical results corresponding to Eq.（5） have been carried out to show that the numerical results have a better exhibition of nonlinear characteristics. Wavelet analysis method has been adopted to investigate the time-frequency energy spectrum of simulation freak waves and the results reveal strong nonlinear interaction enables energy to be transferred to high harmonics during the progress of its formation. Varying water depth can enhance the nonlinear interaction, making much more energy be transferred to high harmonics and freak waves with higher asymmetry be generated.

基于FREAK特征的快速景象匹配

针对传统基于局部不变特征的景象匹配方法处理时间过长的问题,提出基于FREAK局部不变特征的快速景象配准算法。首先提出FAST-Difference特征点提取方法,分别提取参考图像和待配准图像中的特征点;接着计算其FREAK描述符,生成特征向量;随后利用级联匹配计算特征向量之间的汉明距离,提取出匹配的特征点对;最后利用RANSAC算法剔除误匹配点,利用最小二乘法估算出两幅图像之间的空间几何变换参数,实现两幅图像的配准。FAST-Difference相比以往的特征点检测方法速度更快;FREAK描述符与人类视网膜相似的结构提升了算法的时间性能和鲁棒性;使用扫视搜索,大大加速了匹配过程。实验证明相对于SIFT、SURF算法,本文算法不仅对于各种变换具有更好的鲁棒性,而且处理时间大大缩短,实现了景象匹配的实时处理。

基于FAST角点和FREAK描述符改进的无人机景象匹配算法

在无基准图的无人机返航过程中,实时图和航路点的景象匹配是无人机返航成功的关键。为提高景象匹配的实时性和鲁棒性,提出了基于加速分割检测特征(FAST)角点检测和快速视网膜关键点(FREAK)描述符的无人机景象匹配算法。首先,针对FAST角点检测方法的无尺度不变性、特征点数量冗余的缺点进行了改进;接着,对FREAK二进制描述符进行简化,以提高匹配速度;然后,采用K近邻比值法和RANSAC方法进行特征的初匹配和精匹配,并建立定位模型,从而获得航路点与无人机当前位置的实际距离和方位信息;最后,对算法的各项性能做实验验证。所提出的算法定位方向偏差在1°以内,像面距离偏差稳定在0.6 pixel,运行时间0.43 s,远小于尺度不变特征转换(SIFT)和加速鲁棒特征(SURF)算法的处理时间。在尺度变换和噪声等条件变化的情况下,相比SIFT和SURF等算法,所提算法取得了较好的正确匹配率,具有更好的鲁棒性。实验结果表明所提出的算法鲁棒性好,运算速度快,尤其在视角变换方面表现优秀,更适合无人机视觉辅助导航。

基于改进FREAK的图像匹配算法

为提升图像匹配算法的实时性和鲁棒性,提出一种基于改进FREAK的特征点匹配算法。将经典FREAK算法的8层视网膜模型简化为5层,根据贪婪搜索算法选取64组感受野点对,以在减少运算开销的同时尽量保留有效的点对信息。在此基础上,设计一种具有方向不变性的LBP算法对每个感受野进行编码,从而提高描述符的区分度。实验结果表明,与FREAK、BRISK等算法相比,该算法具有最小的描述符尺寸,且在多数场景下,其运算更快,精度更高,更适合光照变化复杂的环境。

一种改进的FREAK算法的图像特征点匹配

由于FREAK算法存在不具备尺度不变性的缺陷,特征点匹配策略单一容易出现匹配效果不理想的问题,鉴于SIFT和RANSAC算法思想,本文提出一种改进的FREAK算法:SFREAK(SIFT and FREAK).首先,生成高斯差分金字塔图像,使检测出来的特征点具有尺度不变性;然后采用FREAK描述符对特征点进行描述,获得二进制描述子;最后,在特征点匹配过程中通过Hamming距离匹配进行粗匹配,并结合RANSAC算法对匹配点对进行提纯,实现两幅图像的特征点匹配.实验结果表明,本文提出的改进算法有效解决了FREAK不具备尺度不变性的缺陷,在图像发生尺度变化时,SFREAK算法特征点匹配准确率达到95.7%,相比于FREAK提高了61.9%.另外,本文改进的算法与传统SIFT、FREAK算法相比,表现出更好的鲁棒性.

FREAK和改进的RANSAC算法在影像匹配中的应用

针对倾斜摄影中包含多个视角的影像而使匹配速度较慢的问题,提出了一种Speeded up robust features(SURF)、Fast Retina Keypoint(FREAK)和改进的Random Sample Consensus(RANSAC)算法相结合的影像匹配算法。首先利用SURF算法提取出稳健的特征点,然后利用FREAK描述符对特征点进行描述,在影像匹配阶段首先进行预匹配,然后用改进的RANSAC算法剔除错误匹配点对。经实验比较分析后得出:该算法在效率和匹配精度上都有较好的鲁棒性。

Efficient Focusing Models for Generation of Freak Waves

Four focusing models for generation of freak waves are presented. An extreme wave focusing model is presented on the basis of the enhanced High-Order Spectral （HOS） method and the importance of the nonlinear wave-wave interaction is evaluated by comparison of the calculated results with experimental and theoretical data. Based on the modification of the Longuet-Higgins model, four wave models for generation of freak waves （a. extreme wave model ＋ random wave model; b. extreme wave model ＋ regular wave model; e. phase interval modulation wave focusing model; d. number modulation wave focusing model with the same phase） are proposed. By use of different energy distribution techniques in the four models, freak wave events are obtained with different Hmax/Hs in finite space and time.

Numerical simulation and mechanism analysis of freak waves

A numerical wave model based on the modified fourth-order nonlinear Schroe dinger equation （mNLSE） in deep water was developed to simulate the formation of freak waves and a standard split - step, pseudo-spectral method was used to solve the equation. The validation of the model is firstly verified, then the simulation of freak waves was performed by changing sideband conditions, and the variation of wave energy was also analyzed in the evolution. The results indicate that Benjamin - Feir instability （ sideband instability） is an important mechanism for freak wave formation.

An experimental and numerical study of the freak wave speed

The propagation speed is one of the most important characteristics for describing freak waves. The research of freak wave speed is not only helpful for understanding the generation mechanism and evolution process of freak waves, but also applicable to the prediction. A stable and accurate method is proposed for the cal- culation of the freak wave speed, in which physical model tests are carried out to measure the motion of the largest wave crest along the wave tank. The linear regression relationship between the spatial position of the largest wave crest and instantaneous moment is established to calculate the speed of totally 248 cases of experimental freak waves and 312 supplementary cases of numerical freak waves. Based on the calculate results, a semitheoretical and semiempirical formula is proposed by using a regression analysis method to predict the speed of the freak wave, and the nonlinear characteristic of the freak wave speed is also investi- gated.