可定位图像采集与检索方法研究

图像是地理信息系统表达中的一种重要数据源,随着应用领域的扩展,行业对图像的需求不再局限于其可视化表达,而是更注重于地理位置等多信息的集成。为此,提出集图像内容、图像元数据、位置特征以及描述信息于一体的可定位图像表达模型,通过Android平台实现图像信息、定位信息、方位信息以及地址描述等要素的实时获取,并从效率、通用性以及可扩展性方面对模型组织方式进行衡量,最终以EXIF图像格式对上述获取的参数进行组织存储,形成可定位图像。针对可定位图像具有的位置以及方位特性,采用规则网格对其位置特征构建空间索引,同时按方位值对方位特征进行归类索引,在此基础上,对福州市区采集的图像进行检索实验,实验结果证明基于位置的图像空间索引方法是可行的,能快速检索到图像。     

基于遮挡区域建模和目标运动估计的动态遮挡规避方法

对于运动视觉目标,如何对遮挡区域进行规避是视觉领域一个具有挑战性的问题.本文提出了一种新颖的基于运动视觉目标深度图像利用遮挡信息实现动态遮挡规避的方法.该方法主要利用遮挡区域最佳观测方位模型和视觉目标运动估计方程,通过合理规划摄像机的观测方位逐渐完成对遮挡区域的观测.主要贡献在于：1）提出了深度图像遮挡边界中关键点的概念,利用其构建关键线段对遮挡区域进行快速建模;2）基于关键线段和遮挡区域建模结果,提出了一种构建遮挡区域最佳观测方位模型的方法;3）提出一种混合曲率特征,通过计算深度图像对应的混合曲率矩阵,增加了图像匹配过程中提取特征点的数量,有利于准确估计视觉目标的运动.实验结果验证了所提方法的可行性和有效性.     

基于力直方图的遥感图像中目标群间方位关系的建立

针对目标群间方位关系的建立,采用力直方图法,对所要处理的图像中目标群做一次全方位扫描,建立数据结构,可将各目标用于力直方图求解的有用信息一次取出,并建立目标群中各目标之间方位关系分布结构表示。最后,用多幅遥感图像验证了方法的可行性,证明了力的直方图法是一种用于表示目标间方位关系的有效方法。该方法一次可对多个目标进行处理,不必重复扫描,减少了计算量,便于对多个目标的把握,有效的掌握了目标群的整体分布特性,满足了同时对多目标图像处理的需求。     

基于图像特征的合成孔径声纳声图像拼接研究

利用合成孔径声纳(SAS)声图方位向线分辨率恒定的特点,提出了基于图像特征的SAS声图拼接方法,其关键技术是使用SURF算子和RANSAC算法。方法使用SURF算子提取SAS声图的特征点,使用RANSAC算法对SAS声图进行精匹配处理。对SAS实测图像进行了处理,结果证明了该方法的有效性。     

长度归一化扫描的合成孔径雷达图像船舶尾迹恒虚警率检测算法

船舶尾迹检测作为近年来SAR图像海洋应用的一个重要方面而受到关注。本文提出了一种基于归一化扫描的SAR图像船舶尾迹检测算法。由于船舶的距离向运动分量会导致船舶在SAR图像上存在方位向偏移,因此算法首先在船舶的方位向一定距离内逐点进行360度归一化扫描得到扫描域,然后在扫描域内以概率模型进行统计建模实现了CFAR检测,进而对检测结果进行形态学处理,最后对扫描域进行K-均值聚类。反演扫描域聚类后的类心得到带端点目标尾迹及方位偏移量,同时自动提取出船舶的航速。利用COSMO-SkyMed数据对本文所提出的方法进行实验验证,实验结果表明本文的算法可以快速有效地进行尾迹检测和船舶速度反演。     

应用血管内超声与X射线造影图像融合的血管三维重建

针对X射线血管造影和血管内超声各自显示血管形态的局限以及互补性问题,提出将两种数据进行融合,准确重建血管的方法。首先从在超声导管回撤路径起点拍摄的两个角度的造影图像对中提取出导管并进行三维重建。然后采用基于snake模型的半自动方法从各帧超声图像中提取出血管壁的内外膜边缘。最后,在Frenet-Serret标架中确定各相邻帧超声图像间的相对方位后,利用非迭代的统计优化方法确定各帧超声图像沿导管轴向的绝对方位,完成两种数据的融合。采用临床数据的实验结果验证了算法的可行性,并分析了可能存在的误差和计算成本。     

基于图像处理的移动机器人对目标的识别和定位

为实现移动机器人对远距离初始焊位的导引工作,介绍了-种简单有效的方法;通过图像处理计算出主动目标图像特征的质心和方位点来实现对目标的锁定,并利用运动学建模确立了主动目标从图像到实际工作空间的映射关系,从而得到目标实际的位置和方位.     

罗经光学方位仪模拟器实时景象生成系统开发设计

模拟视景图像是罗经光学方位仪模拟器中的一个重要组成部分和设计难点,论文研究了罗经光学方位仪模拟器实时景象生成系统中图像发生器、视景驱动软件、地形数据库和三维活动目标数据库等组成、特点和开发设计方法。     

基于二次曲线的线阵相机标定技术

针对绕固定轴旋转的线阵相机,提出一种基于二次曲线的相机标定方法。该方法只需要相机在2个或更多不同的方位拍摄图像,通过相机的转动在每个方位拍摄多帧图像。靶标采用一个包含3个或更多二次曲线的平板。相机和靶标可以自由移动,不需要知道运动参数。通过坐标变换,把每个方位所拍摄的多帧线阵图像排列成面阵图像。从面阵图像中提取二次曲线作为标定基元,以此简化基元对应问题。仿真实验结果表明,该方法精度较高,鲁棒性较好。     

全极化合成孔径雷达海浪遥感方法

提出了一种利用全极化真实孔径雷达测量方位方向和距离方向海浪斜率的新方法,该方法不同于极化方向角估计海浪方位向斜率方法,利用两种线极化图像信息将有关弱极化项剔除(流体动力学调制),获得仅包含倾斜项和极化调制项的雷达成像公式。利用机载SAR取代真实孔径雷达进行海浪反演,获得了与浮标致的结果。另外,在速度聚束模式情况下,推导了双极化合成孔径雷达图像谱同海浪谱的非线性变换关系。     

Closed-loop tool face control with the bit off-bottom

In this paper, we develop three methods to achieve reliable closed-loop, tool face control for directional drilling operations. This is a necessary step to achieve closed-loop, automated directional guidance. Our algorithms combine existing industry top-drive controllers with new control approaches. The torsional model we use for the drill string has been field validated and takes into account the Coulomb friction between the drill string and the borehole. These distributed friction terms are either assumed known (or measured) or can be estimated using a state-observer. In this work, we improve such a state-observer to obtain an estimation of the tool face orientation in real-time. We then propose different approaches to control the tool face. The first method is based on a feed-forward control law. It uses the flatness of the model and the estimation of the orientation to generate an admissible trajectory which is then tracked. In the second procedure, we require a stable rotation off-bottom before smoothly changing the reference to zero to stop bit rotation. This change of reference induces a change of orientation that can be estimated and finally compensated by repeating the procedure. Finally, the last method uses a series of trapezoidal setpoint inputs – bumps – to calculate the change in downhole tool face per change in surface orientation before arriving at the correct tool face after three iterations. These three algorithms are illustrated in simulations of field scenarios and their effectiveness and limitations, depending on the reliability and availability of downhole orientation data, are discussed.     

In-plane face orientation estimation in still images

This paper addresses a fine in-plane (roll) face orientation estimation for a perspective face analysis algorithm that requires normalized frontal faces. As most of the face analysers (e.g., gender, expression, and recognition) need frontal up-right faces, there is a clear need for the precise roll estimation, as precise face normalization has an important role in classification methods. The in-plane orientation estimation algorithm is constructed on top of regular Viola-Jones frontal face detector. When a face is detected for the first time, it is rotated with respect to the face origin to find the boundaries of the detection. Mean value of these angles is said to be the measurement of the in-plane rotation of the face. Since we only need a face detection algorithm, the proposed method can work effectively on very small sized faces where traditional landmark (eye, mouth) or planar detection based estimations fail. Experiments on controlled and unconstrained large-scale datasets (CMU Rotated, YouTube, Boston University Face Tracking, Caltech, FG-NET Aging, BioID and Manchester Talking-Face) showed that the proposed method is robust to various settings for in-plane face orientation estimation in terms of RMSE and MAE. We achieved less than ±3.5 ^° mean absolute error for roll estimation which proves that the accuracy of the proposed method is comparable to that of the state-of-the-art tracking based approaches for the roll estimation.     

一种人脸姿势判别与正脸合成的新方法

人脸深度旋转是人脸识别领域的一个瓶颈问题.首先探讨了不同方向、尺度的Gabor滤波器对于判别不同朝向的人脸姿势的性能,然后提出了一个基于Gabor滤波和分数幂多项式核Fisher判别准则的人脸姿势判别方法,最后利用改进的点点对应算法和线性物体类的原理构造正脸合成的算法.实验结果表明,新提出的姿势判别和合成方法是有效的.     

Harmonic point cloud orientation

In this work we propose a new method for estimating the normal orientation of unorganized point clouds. Consistent assignment of normal orientation is a challenging task in the presence of sharp features, nearby surface sheets, noise, undersampling, and missing data. Existing approaches, which consider local geometric properties often fail when operating on such point clouds as local neighborhood measures inherently face issues of robustness. Our approach circumvents these issues by orienting normals based on globally smooth functions defined on point clouds with measures that depend only on single points. More specifically, we consider harmonic functions, or functions which lie in the kernel of the point cloud Laplace-Beltrami operator. Each harmonic function in the set is used to define a gradient field over the point cloud. The problem of normal orientation is then cast as an assignment of cross-product ordering between gradient fields. Global smoothness ensures a highly consistent orientation, rendering our method extremely robust in the presence of imperfect point clouds.     

Multi-view face identification and pose estimation using B-spline interpolation

The available face views in the training set are mostly limited. In this paper, we present a view interpolation method using nonlinear B-spline on face manifolds. Two models, the inner–outer ellipse model and the moment of inertia model, are developed to estimate the pose orientation. We use the limited view-pose face images to form the pose eigen space. Then, based on these nonlinear manifolds we form a B-spline for each individual. Identification is to compute the shortest Euclidean distance from a given test view to the nearest point within one of these B-splines. Once the test view is classified as a familiar individual in the training set, not only can the individual be identified, but also the pose angle can be estimated. Experimental results show that B-spline interpolation can achieve a recognition rate of 95%.     

The Big Data Analysis on the Camera-based Face Image in Surveillance Cameras*

In the Big-Data era, currently how to automatically realize acquisition, refining and fast retrieval of the target information in a surveillance video has become an urgent demand in the public security video surveillance field. This paper proposes a new gun-dome camera cooperative system, which solves the above problem partly. The system adopts a master-slave static panorama-variable view dualcamera cooperative video-monitoring system. In this dual-camera system the gun camera static camera） with a wide viewing -angle lenses is in charge of the pedestrian detection and the dome camera can maneuver its focus and cradle orientation to get the clear and enlarged face images. In the proposed architecture, Deformable Part Model (DPM) method realizes real-time detection of pedestrians. The look-up table method is proved feasible in a dual-camera cooperative calibration procedure, while the depth information of the moving target changes slightly. As respect to the face detection, the deep learning architecture is exploited and proves its effectiveness. Moreover, we utilize the Haar-Like feature and LQV classifier to execute the frontal face image capture. The experimental results show the effectiveness and efficiency of the dual-camera system in close-up face image acquisition.     

Reliable facial expression recognition for multi-scale images using weber local binary image based cosine transform features

Multimedia Tools and Applications - Accurate recognition of facial expression is a challenging problem especially from multi-scale and multi orientation face images. In this article, we propose a...     

一个人脸识别系统的设计与实现

设计了一个在区分人脸朝向的基础上,再进行人脸辨别的识别系统。把BP网络用于人脸朝向的识别,并且用一组单输出的子网络代替原来的BP网络,提高了人脸朝向的识别率。在人脸朝向的分类基础上,用主元分析法进行人脸识别。实验结果表明此系统提高了人脸的识别率。     

基于Gabor滤波特征和支持向量机的人脸检测

人脸检测是人脸识别与图像及视频检索的一项重要任务。论文提出了一种基于Gabor滤波特征和支持向量机的正面人脸检测方法。算法首先利用了Gabor滤波器的良好的空间位置与方向的选择特性,采用了四种方向的Gabor滤波器提取人脸样本图像特征并用PCA方法对特征降维,然后用已降维的特征训练支持向量机分类器。最后应用SVM分类检测人脸。实验结果证明该方法行是十分有效的。     

基于方向边缘幅值的尺度块LBP人脸识别

针对方向边缘幅值模式（POEM）忽略了块与块之间的像素问题,提出一种基于方向边缘幅值的尺度块LBP人脸识别方法。该方法首先用梯度算子提取出人脸的方向图和幅值图,将具有相同量化方向上的幅值累加,然后利用尺度块LBP算子提取每幅累加幅值图的分块直方图特征,并将所有直方图特征串联起来作为最终的识别特征,最后采用WPCA降维方法提高算法的有效性。实验结果表明,该算法的鲁棒性高于其他对比算法,运用降维处理后能以较低的特征维数达到良好的识别性能。