Invariant pattern recognition based on centroids

A new method for pattern recognition that is invariant under changes of position, orientation, intensity, and scale is presented. The centroids of objects provide unique points that are related to the energy distribution. For obtaining more such unique points a conformal transform can be used to rearrange the energy distribution of the object. By means of the conformal transform many different centroids can be produced from the same object. A useful pattern-recognition and object-registration method that yields a position-, rotation-, intensity-, and scale-invariant feature vector based on these centroids can be created.     

Invariant pattern recognition by use of wavelength multiplexing

Rotation-invariant pattern recognition can be achieved with circular-harmonic decomposition. A common problem with such a filter is that, because it is only a single term out of the circular decomposition, it does not contain much of the reference object's energy. Thus, the obtained correlation selectivity is low. This problem is solved by use of wavelength multiplexing. First, different harmonic terms are encoded by different wavelengths, and then they all are added incoherently in the output correlation plane. This process leads to rotation-invariant pattern recognition with a higher discrimination ability.     

Lossless parameterisation of image contour for shape recognition

A parameterised contour that is invariant against affine transformations is a convenient substitution of the object image for shape recognition. Generally, parameterisation needs several representative signals of the image contour to fit the affine transformation model. When the representative signals fail to carry the contour information thoroughly, information loss occurs in the resulting parameterised contour, and so the accuracy of shape recognition may deteriorate. Synthesised feature signals are shown, which represent that an image contour without information loss can be extracted with partial Fourier synthesis or partial cosine synthesis. Lossless parameterisation of the image contour is obtained by substituting the synthesised feature signals into the affine invariant function. Experimental results verify its representative, affine invariance and recognition rate in shape recognition. The results are compared with those by partial wavelet synthesis, which has insignificant information loss.     

Invariant pattern recognition by use of a spatial code division multiplexing approach

Invariant pattern recognition can be achieved by use of harmonic decomposition, for example circular harmonics are used for rotation invariant recognition. A common problem with such methods is that often only a single term of the harmonic decomposition is used, and it does not contain a sufficient amount of the reference energy. Thus discrimination capability is limited, especially in the presence of noise or other disturbances. By using several terms of the harmonic decomposition together this problem can be solved; this can be achieved by the use of code division filter multiplexing. Several harmonic terms are encoded onto a single filter, and the signal is simultaneously correlated with all of them, hence producing enhanced discrimination capabilities. Here two methods are suggested for such encoding. The first involves multiplexing the filters in the Fourier plane, while the second involves multiplexing in the image plane.     

一种基于边界特征的耳廓提取新方法

在实际的人耳识别系统中,人耳的准确定位是影响识别率的一个重要因素.根据外耳及其所在位置的特征,提出了一种从侧脸图像上准确定位并提取出人耳的新方法.该方法首先对侧脸图像进行阈值化和差分处理,然后将外耳边缘轮廓分为3个区域,利用阈值图像和差分图像以及3个区域的不同走向分别跟踪得到整个外耳轮廓线,并从侧脸图像中提取出外耳图像.该方法运算量小,定位精确,得出的外耳轮廓图像清晰完整无任何干扰,而且定位成功率高.     

人脸识别的一种新的特征提取方法

针对人脸识别中判别特征的提取问题,本文提出了一种新的人脸识别算法—扩展保局投影（ELPP）。普通保局投影（LPP）在构建权图时侧重保持样本的局部结构,属于无监督学习算法。扩展保局投影在保局投影的基础上进行扩展,通过引入可调因子,在保持人脸图像局部流形结构的同时考虑样本的类别信息,从而充分提取样本的判别特征。本文采用最小近邻分类器估算识别率。在Yale人脸库以及AT＆T人脸库的测试结果表明,在姿态、光照、表情、训练样本数目变化的情况下,ELPP都具有较好的识别率。     

融合分数维特征的水岸彩色图像水岸界线识别

针对复杂的水岸景物彩色图像水岸界线识别,提出了融合多特征的分数维方法。该方法用提取边界来降低维数;再将横向细长窗口内的 Lebesgue 测度(μmin =260)和水岸界线形状位置等特征作为限定条件,只对符合条件点应用基于 ε-blanket 方法的覆盖技术进行迭代计算,找到分数维数最接近 1 的窗口;用经过部分改造的最小二乘法拟合窗口内符合条件点即为水岸界线。这样可以大幅度减少计算量,需迭代计算像素点数量仅为原算法的 0.88%。对多种自然条件下水岸图像 1026 幅进行识别,完全识别率达 85.3%,反应出该方法的鲁棒性与实用性。     

Optimum design of dome contour for filament-wound composite pressure vessels based on a shape factor

Filament-wound composite pressure vessels are an important type of high-pressure container that is widely applied in the commercial and aerospace industries. This study investigates the optimum design of dome contours for filament-wound composite pressure vessels, subjected to geometrical limitations, winding condition, and the Tsai–Wu failure criterion and maximizing shape factor, the feasible direction method being employed. An actual design example, presented by Fukunaga et al. [19], is adopted to study the optimum dome contour using the present method. Results reveal that the dome contours using the present method, Fukunaga et al.’s method and the netting method can be approximated using elliptic curves, and that the depth is the major parameter for optimizing the design of dome contour, and the dome, designed using the present method has stronger structure and greater internal volume than those designed using other approaches. Results reveal that the present method is usable for the optimum design of dome contours for filament-wound composite pressure vessels.     

Emotion recognition based on facial components

Machine analysis of facial emotion recognition is a challenging and an innovative research topic in human–computer interaction. Though a face displays different facial expressions, which can be immediately recognized by human eyes, it is very hard for a computer to extract and use the information content from these expressions. This paper proposes an approach for emotion recognition based on facial components. The local features are extracted in each frame using Gabor wavelets with selected scales and orientations. These features are passed on to an ensemble classifier for detecting the location of face region. From the signature of each pixel on the face, the eye and the mouth regions are detected using the ensemble classifier. The eye and the mouth features are extracted using normalized semi-local binary patterns. The multiclass Adaboost algorithm is used to select and classify these discriminative features for recognizing the emotion of the face. The developed methods are deployed on the RML, CK and CMU-MIT databases, and they exhibit significant performance improvement owing to their novel features when compared with the existing techniques.     

Shift- and scale-invariant recognition of contour objects with logarithmic radial harmonic filters

The phase-only logarithmic radial harmonic (LRH) filter has been shown to be suitable for scale-invariant block object recognition. However, an important set of objects is the collection of contour functions that results from a digital edge extraction of the original block objects. These contour functions have a constant width that is independent of the scale of the original object. Therefore, since the energy of the contour objects decreases more slowly with the scale factor than does the energy of the block objects, the phase-only LRH filter has difficulties in the recognition tasks when these contour objects are used. We propose a modified LRH filter that permits the realization of a shift- and scale-invariant optical recognition of contour objects. The modified LRH filter is a complex filter that compensates the energy variation resulting from the scaling of contour objects. Optical results validate the theory and show the utility of the newly proposed method.     

基于超像素的联合能量主动轮廓CT图像分割方法

为解决医学CT图像主动轮廓分割方法中对初始轮廓敏感的问题,提出一种基于超像素和卷积神经网络的人体器官CT图像联合能量函数主动轮廓分割方法。该方法首先基于超像素分割对CT图像进行超像素网格化,并通过卷积神经网络进行超像素分类确定边缘超像素;然后提取边缘超像素的种子点组成初始轮廓;最后在提取的初始轮廓基础上,通过求解本文提出的综合能量函数最小值实现人体器官分割。实验结果表明,本文方法与先进的U-Net方法相比平均Dice系数提高5%,为临床CT图像病变诊断提供理论基础和新的解决方案。     

改进测地线活动轮廓模型的肝脏超声图像分割

针对超声医学图像的特点,系统地发展和应用测地线活动轮廓模型技术进行肝脏超声图像的边缘提取,提出一种改进的测地线活动轮廓模型.首先把总变分和各向异性扩散方程的有机结合构造出新的边缘停止函数;其次,提出三种边缘停止函数,通过实验比较得出最佳模型.新模型能有效地保留目标信息,消除噪音,增大图像梯度值,具有更强的收敛效果.实验表明,该方法能更加准确地提取弱边缘,达到更理想的分割效果.     

Modernization of a conceptual model of a data bank for physicotechnical effects based on contemporary information technology

Requirements for a conceptual model of a data bank for physicomechanical effects and methods for achieving them are studied. On the basis of presenting converters as an information system, a method for modelling them is suggested based on the theory of an energy-information model, parametric structural schemes, SADT-technology, and an objectively orientated paradigm. Advantages of the conceptual model developed are discussed. __________ Translated from Izmeritel’naya Tekhnika, No. 1, pp. 7–9, January, 2007.     

基于GNSA多尺度模型的人脸识别

考虑近似图像信息和细节图像信息,提出了 GNSA 多尺度模型。该模型利用具有 1 个隐含层和 50 个隐单元的神经网络建立不同尺度图像间的映射关系;使用反向传播算法训练神经网络,确定这种映射关系;根据该映射关系由低分辨力图像估计高分辨力图像。采用亮度相似性和对比度相似性量化估计图像与目标图像间的相似程度。实验表明,以该模型分析得到的两种相似性分别为 89.907%和 96.196%;以该模型为基础的人脸识别系统对光照具有很好的鲁棒性。     

基于数据融合的红外目标识别

为了获得良好的红外目标识别性能,综合应用了图像处理、模式识别和数据融合领域内的新技术.采用了神经网络和证据理论集成的数据融合方法进行目标识别的数据融合.根据LVQ神经网络在目标识别领域内应用特点,构造了基于证据理论的基本概率赋值函数.对此目标识别技术进行了测试,结果表明,采用此技术后的识别的可信度得到了较大提高.     

基于不变特征的标志识别技术研究

标志识别是近距离空间交会对接中的一项重要的关键技术.在空间交会对接过程中,太阳和月球等各种天体以及飞行器上零部件的辐射与反射都会影响CCD的成像,形成大量杂散光干扰.为此,本文首先设计了具有不变特征的标志系统,然后根据标志系统固有的不变特征,详细设计了左右像点匹配、远场像平面不变特征识别、近场像平面预测识别以及像空间模型匹配全等识别等一套完整算法,可以保证彻底消除杂散光的干扰,正确可靠地完成各个标志灯像点的辨识.模拟图像仿真实验和真实图像仿真实验结果验证了算法的正确性和可靠性.     

步态能量图和KFDA的步态识别研究

为了有效地获取步态连续性的动态特征,快速准确地进行身份识别.特提出了一个基于步态能量图(Gait Energy Image,GEI)和核Fisher判别分析(Kernel-based Fisher Discrimination Analysis,KFDA)的分类识别算法.算法首先以步态能量图(GEI)按列向量作为输入,求得最优子空间W_(opt)和α_(opt).利用提取步态能量图(GEI)的步态信息向量计算在α_(opt)上的投影,并计算其投影轨迹.在分类阶段,采用最近邻分类器(Nearest neighbor classifier).最终在中科院自动化研究所CASIA B步态数据库上进行实验,对比多项式、高斯径向基核函数和其他四种算法的结果显示,本文算法取得了较高的识别率.     

基于Hausdorff距离的工件焊缝接头类型识别

提出了一种以Hausdorff距离为度量准则的工件焊缝接头类型识别方法.通过激光结构光视觉方法,CCD摄像机采集被焊工件焊缝接头的激光图像,并提取出激光条纹中心线.根据焊缝接头图像特点,提出6种焊缝接头条纹形状基准模型.以Hausdorff距离度量激光条纹图形与基准模型库之间的匹配程度,当条纹形状与基准模型的距离具有最小容差时,认为与相对应的焊缝接头类型相匹配.为实现算法加速,采用Hough变换和Hotelling变换来标准化图形,并利用图像特征对搜索空间进行了降维.实验结果验证了该识别方法的有效性.