基于图像统计特性的商标图像检索

提出了单元图像和单元特征的概念,并在不同层次上从不同角度对图像进行多重描述。根据这些描述用欧氏距离来度量图像间的相似性。实验结果表明,本方法具有良好的检索效果。     

Real-time rendering algorithm based on a hybrid rendering scheme

A real-time rendering algorithm based on a hybrid rendering scheme is presented. The key frames of the virtual scene are generated with a geometry-based approach, then a non-linear transformation to all visible points of each frame near the current viewpoint is performed to reveal their coplanar characteristics and a hierarchical polygon approximation of the local scene is set up. Any intermediate image between the two key frames could be obtained by re-projecting these polygons onto the view plane related to the current viewpoint. The gaps that might appear on the intermediate image are filled through bi-directional backward image warping. Experimental results show that our approach is insensitive to the complexity of the scene and has a satisfactory performance.     

Mechanical performance and texture characteristic of an IF steel containing Nb and Ti by double cold rolling

Single cold rolling and double cold rolling were applied to hot rolled strips with different reduction ratios. The evolutions of { 100}, { 111} and Goss face texture during double rolling were investigated by comparing the orientation distribution function (ODF) of the double rolled sample with that of the single rolled one. The double cold rolling texture is characterized by a higher γ-texture and a lower α-texture, and the { 111}〈112〉 component is improved remarkably. Based on the TEM observation and mechanical properties test, it is found that the reduction ratio assignment significantly affects the texture variation, as-annealing microstructures, and properties of the double cold rolled samples. These results may provide a theoretical guide for the industrial production of double cold rolled IF steel.     

结合Nystrm逼近的图半监督纹理图像分割

针对半监督学习算法在图像分割中的应用,提出了一种基于流形插值的半监督图像分割方法。该方法将分类问题看作一个流形上的函数的插值问题,通过优化某些系数来更好地拟合数据。该算法采用稀疏图可解决大规模矩阵特征值和特征向量的求解。但是,对于图像分割来说,构造稀疏图的运算时间较长,针对这一问题,提出采用Nystrm逼近方法来降低计算复杂度。合成纹理图像分割结果验证了该算法可获得良好的分割质量,结合Nystrm逼近方法在保证分割质量的前提下从很大程度上提高了计算效率。     

海洋遥感图像中基于纹理的海面油膜识别

基于海面纹理特征,利用Borda算法进行了航空遥感图像中油膜和非油膜的识别研究。首先,对航空遥感图像进行纹理特征分析和提取:二值化处理并且提取谱相关性、方向粒度和条状粒度的纹理特征值;然后,建立海面油膜数据库,包括原始图像、特征向量,以及图像和纹理特征间的映射表;最后引入Borda算法,根据被测图像的特征来识别其是否含有油膜。实验结果表明,本模型能有效的识别油膜,能为大范围海面溢油检测奠定研究基础。     

Textural changes in metallurgical coke prepared with polyethylene

The effect of high-density polyethylene (HDPE) on the textural features of experimental coke was investigated using polarized-light optical microscopy and wavelet-based image analysis. Metallurgical coke samples were prepared in a laboratory-scale furnace with 2.5%, 5.0%, 7.5%, 10.0%, and 12.5% HDPE by mass, and one sample was prepared by 100% coal. The amounts and distribution of textures (isotropic, mosaic and banded) and pores were obtained. The calculations reveal that the addition of HDPE results in a decrease of mosaic texture and an increase of isotropic texture. Ethylene formed from the decomposition of HDPE is considered as a probable reason for the texture modifications. The approach used in this study can be applied to indirect evaluation for the reactivity and strength of coke.     

鱼糜及鱼糜制品质构的评价方法

总结了国内外有关鱼糜及鱼糜制品质构的仪器测定法和感官评价法.仪器测定法通过测定鱼糜及鱼糜制品的凝胶性能、质构特性参数、弹性模量等来评价鱼糜及鱼糜制品的质构;感官评价法主要利用折叠实验和咀嚼实验评价鱼糜及鱼糜制品的质构.最后,对该两种评价方法做了对比分析.     

CSP流程下退火加热制度对深冲板织构和性能的影响

以包钢CSP流程热轧板为基料的SPCD冷轧板为试验材料,采用单台阶和双台阶两种退火加热制度进行实验室试验,对比分析了不同规程下的再结晶织构、组织和性能特点.结果表明,双台阶退火工艺对应的γ织构、α织构上的{111}〈110〉成分均有所提高,但是双台阶工艺下的Δr值偏大.分析认为,双台阶和单台阶在回复阶段上有所不同,不利取向组分形核受到了抑制,最终导致有利织构含量上的差别.在双台阶工艺下,{110}〈110〉织构强度的提高导致Δr值偏大.     

综合颜色和纹理特征的粒子滤波人脸跟踪算法

提出了一种综合颜色和纹理特征的粒子滤波人脸跟踪算法.该方法利用粒子滤波能有效处理非线性非高斯过程和融合目标人脸多种测量信息的特性,针对人脸特征对环境变化的不同鲁棒性,综合加权颜色直方图和旋转复合小波进行人脸特征描述,将颜色和纹理特征融合到粒子滤波跟踪框架中.实验结果表明了该算法的鲁棒性、精确性和灵活性,与基于单一特征的粒子滤波跟踪方法相比,该算法稳健而有效,且对现实场景下人脸的跟踪效果准确.     

基于改进Hough变换的刀具磨损状态监测

针对传统Hough变换的参数空间所需存储量大、运算量大、不利于实现实时监测等缺点,提出了一种基于随机Hough变换的改进Hough变换算法,并将该方法用于工件表面纹理图像的处理,通过对工件表面的二值边缘图像进行改进Hough变换,提取直线段平均长度及直线段与切削速度方向的夹角作为特征参数,实现对刀具磨损状态的判断.算法分析和实验结果表明,该算法所需存储空间少,计算量小;提取的特征参数与刀具磨损状态之间存在密切联系,根据特征参数的变化规律可实现对刀具磨损状态的监测.