基于透视不变量的平面曲线的识别

提出一种快速有效的平面曲线的识别方法,在这种方法中,我们从单幅图象中提取表征曲线形状的描述子,该描述子由曲线上透视不变点之间的射影距离不变量构成,且不受物体和摄象机之间的位置、方向以及摄象机物理参数的影响。因此这种识别算法不受透视畸变的限制,本文用一组起初的平面物体进行了识别实验,结果表明这些形状描述 平面 曲线的识别有很好的性能。     

面阵CCD航空相机斜视图像的几何校正

为增大机载光电侦察系统的视场角,航空相机多采用扫描方式对地摄影成像,文章针对面阵CCD航空相机斜视摄影引起的图像变形进行校正。由于该类相机与载机之间存在方位和俯仰两轴运动,文章同时考虑相机相对载机的旋转角度和载机相对地面的姿态角,建立了更符合该类相机工作状态的六姿态角投影校正模型,推导出同一地物在畸变图像和标准图像上的像素坐标变换关系,采用双线性插值算法对坐标变换后的像素灰度值进行重采样,对航空图像进行自动校正。最后,对某机场的航拍图像进行校正实验,并与基于地面控制点的多项式校正方法以及基于畸变图像和参考图像配准的校正方法进行比较。实验结果表明,六姿态角投影校正模型能够取得比较高的校正精度(可达像素级),当姿态角的测量精度为3′,图像大小为512pixel×512pixel时,图像校正的均方根误差为1.174 7pixel。该方法不需要提供参考图像和野外采集地面控制点数据,便于工程实现,基本满足航空图像处理的稳定可靠、精度高、实时性强等要求。     

Video orbits of the projective group a simple approach tofeatureless estimation of parameters

We present direct featureless methods for estimating the eight parameters of an "exact" projective (homographic) coordinate transformation to register pairs of images, together with the application of seamlessly combining a plurality of images of the same scene, resulting in a single image (or new image sequence) of greater resolution or spatial extent. The approach is "exact" for two cases of static scenes: (1) images taken from the same location of an arbitrary three-dimensional (3-D) scene, with a camera that is free to pan, tilt, rotate about its optical axis, and zoom, or (2) images of a flat scene taken from arbitrary locations. The featureless projective approach generalizes interframe camera motion estimation methods that have previously used a camera model (which lacks the degrees of freedom to "exactly" characterize such phenomena as camera pan and tilt) and/or which have relied upon finding points of correspondence between the image frames. The featureless projective approach, which operates directly on the image pixels, is shown to be superior in accuracy and the ability to enhance the resolution. The proposed methods work well on image data collected from both good-quality and poor-quality video under a wide variety of conditions (sunny, cloudy, day, night). These new fully automatic methods are also shown to be robust to deviations from the assumptions of static scene and no parallax.     

Building Recognition Based on Geometric Model in FLIR Image Sequences

In this paper, we propose a novel method for building recognition from forward looking infrared (FLIR) image sequences with clutter in the background for automatic target recognition (ATR). In the first phase, dynamic feature space is defined, and camera model and multi-scale characteristic views model that are used to predict model’s features based on 3D object reference model are introduced. In the second phase, the original image is preprocessed using morphological grayscale filters that respond to the size, shape and orientation of object to suppress the background that contains the non-stationary nature and man-made objects of the clutter image. In the following phase, segmentation for the result of image preprocessing is used to obtain regions of interest (ROIs), and features extraction of ROIs and matching retain ROIs that are closest to predicted features. Lastly, the object is identified by fusing the line features. Experiment results show the algorithm can recognize the object from FLIR image sequences with a complicated background.     

智能型目标识别系统

介绍了图像目标识别技术中的图像分割，不变性参数提取和目标分类，利用图像目标的均匀性和相应知识自适应地分割和提取图像目标，被提取的每个图像目标的不变性参数由归一化过程和Ｚｅｒｎｉｋｅ矩提取，并利用ＭＰＮＮ模型将图像目标分类，实验结果该识别系统能识别光照不均匀或复杂背景下的图像目标。     

Textural feature based face recognition for single training images

A novel face recognition algorithm using single training face image is proposed. The algorithm is based on textural features extracted using the 2D log Gabor wavelet. These features are encoded into a binary pattern to form a face template which is used for matching. Experimental results show that on the colour FERET database the accuracy of the proposed algorithm is higher than the local feature analysis (LFA) and correlation filter (CF) based face recognition algorithms even when the number of training images is reduced to one. In comparison with recent single training image based face recognition algorithms, the proposed 2D log Gabor wavelet based algorithm shows an improvement of more than 3% in accuracy.     

Object recognition using multilayer Hopfield neural network

An object recognition approach based on concurrent coarse-and-fine matching using a multilayer Hopfield neural network is presented. The proposed network consists of several cascaded single-layer Hopfield networks, each encoding object features at a distinct resolution, with bidirectional interconnections linking adjacent layers. The interconnection weights between nodes associating adjacent layers are structured to favor node pairs for which model translation and rotation, when viewed at the two corresponding resolutions, are consistent. This interlayer feedback feature of the algorithm reinforces the usual intralayer matching process in the conventional single-layer Hopfield network in order to compute the most consistent model-object match across several resolution levels. The performance of the algorithm is demonstrated for test images containing single objects, and multiple occluded objects. These results are compared with recognition results obtained using a single-layer Hopfield network.     

NCC特征匹配的类脑视觉识别记忆算法

大脑能在较短的时间内以较高的准确率对物体、场景等进行识别;而现有的机器学习算法则可能因图像的微小变化而无法成功识别对象.这主要是因为现有的机器学习算法在识别过程中着重逐层从对象的低级特征提取高级特征,而不能从观察对象的图像中直接提取高级特征.故可建立模型,以Normalized Cross Correlation(NC...     

Projective rectification based on relative modification and size extension for stereo image pairs

For stereo vision applications, projective geometry has proved to be a useful tool for solving the rectification problem without camera calibration. However, the criterion of minimisation for projective rectification must be chosen properly in order to avoid unduly geometric distortion. In this paper, an improved algorithm to minimise the distortion by combining a newly developed projective transform with a properly chosen shearing transform is proposed. The emphasis on low geometric distortion makes this method not only appropriate for 3D reconstruction but also for stereoscopic viewing applications. On the basis of relative modification, this new method contains fewer parameters (6 degrees of freedom) for minimisation, which reduces the processing time, and improves the rectification result. Several different types of image pairs were tested to demonstrate the applicability and reliability of the proposed algorithm visually and quantitatively. Comparisons with other methods are also provided to verify the improvement of this new scheme.     

Preferential Image Segmentation Using Trees of Shapes

A novel preferential image segmentation method is proposed that performs image segmentation and object recognition using mathematical morphologies. The method preferentially segments objects that have intensities and boundaries similar to those of objects in a database of prior images. A tree of shapes is utilized to represent the content distributions in images, and curve matching is applied to compare the boundaries. The algorithm is invariant to contrast change and similarity transformations of translation, rotation and scale. A performance evaluation of the proposed method using a large image dataset is provided. Experimental results show that the proposed approach is promising for applications such as object segmentation and video tracking with cluttered backgrounds.      

3D Tracking and Classification System Using a Monocular Camera

This paper details a 3D tracking and recognition system using a single camera. The system is able to track and classify targets in outdoors and indoors scenarios, as long as they move (at least approximately) on a plane. The system first detects and validates targets and then tracks them in a state-space employing cylindrical models (horizontal and vertical position on the ground, their radius and height) utilising Particle Filters. The tracker fuses visual measurements that utilise the targets’ foreground and colour models. Finally the system classifies the tracked objects based on the visual metrics extracted by our algorithm. We have tested our model in an outdoor setting using humans and automobiles passing through the field of view of the camera at various speeds and distances. The results presented in this paper show the validity our approach.     

一种基于神经网络的立体校正方法

提出了一种简便的立体校正算法。该方法巧妙地运用投影成像原理，构造出一组立体图对在校正前后的对应点，将其作为样本来训练BP神经网络，从而将整个校正关系存储到网络中。利用训练好的神经网络，可以快速有效地实现立体校正。与传统方法相比，该算法的特点是：(1)校正过程中没有利用到摄像机的内外参数，因此无需进行复杂的摄像机定标运算；(2)校正后的图像大小与极点所在位置无关，不会出现校正图像无界的情况。通过实际应用，验证了该方法的正确性和有效性。     

Genetic algorithm for model-based matching of projected images of three-dimensional objects

A novel genetic algorithm (GA) is proposed for searching for the existence of a projective transform which, when applied to the model, results in the best alignment with an unknown 2D edge image. The presence of a valid solution reflects that the latter can be regarded as one of the projected views of the model. On this basis, the identity of an unknown edge image can be deduced by matching it against a set of 3D reference models. To increase the efficiency of the process, a two-pass, coarse-to-fine strategy is adopted. Initially, an unknown image is first classified to a small group of models by matching their outermost boundaries. Next, a fine but slower matching algorithm selects model(s) that share similar internal edge features as the unknown image. In the design of the method, the authors have adopted an alternative projective transform representation that involves fewer parameters and allows constraints to be easily imposed on their dynamic ranges. This effectively lowers the chance of premature convergence and increases the success rate. Experimental results obtained with the proposed scheme are encouraging and demonstrate the feasibility of the approach.     

Three-dimensional face reconstruction from a single image by a coupled RBF network

Reconstruction of a 3-D face model from a single 2-D face image is fundamentally important for face recognition and animation because the 3-D face model is invariant to changes of viewpoint, illumination, background clutter, and occlusions. Given a coupled training set that contains pairs of 2-D faces and the corresponding 3-D faces, we train a novel coupled radial basis function network (C-RBF) to recover the 3-D face model from a single 2-D face image. The C-RBF network explores: 1) the intrinsic representations of 3-D face models and those of 2-D face images; 2) mappings between a 3-D face model and its intrinsic representation; and 3) mappings between a 2-D face image and its intrinsic representation. Since a particular face can be reconstructed by its nearest neighbors, we can assume that the linear combination coefficients for a particular 2-D face image reconstruction are identical to those for the corresponding 3-D face model reconstruction. Therefore, we can reconstruct a 3-D face model by using a single 2-D face image based on the C-RBF network. Extensive experimental results on the BU3D database indicate the effectiveness of the proposed C-RBF network for recovering the 3-D face model from a single 2-D face image.     

基于深度卷积网络的3D人脸重构算法

基于单张人脸图片的3D人脸模型重构,无论是在计算机图形领域还是可见光成像领域都是一个极具挑战性的研究方向,对于人脸识别、人脸成像、人脸动画等实际应用更是具有重要意义。针对目前算法复杂度较高、运算量较大且存在局部最优解和初始化不良等问题,本文提出了一种基于深度卷积神经网络的单张图片向3D人脸自动重构算法。该算法首先基于3D转换模型来提取2D人脸图像的密集信息,然后构建深度卷积神经网络架构、设计总体损失函数,直接学习2D人脸图像从像素到3D坐标的映射,从而实现了3D人脸模型的自动构建。算法对比与仿真实验表明,该算法在3D人脸重建上的归一化平均误差更低,且仅需一张2D人脸图像便可自动重构生成3D人脸模型。所生成的3D人脸模型鲁棒性好,重构准确,完整保留表情细节,并且对不同姿态的人脸也具有较好的重建效果,能够在三维空间中无死角自由呈现,将满足更多实际应用需求。     

Appearance based background subtraction for PTZ cameras

Traditional background subtraction algorithms assume the camera is static and are based on simple per-pixel models of scene appearance. This leads to false detections when the camera moves. While this can sometimes be addressed by online image registration, this approach is prone to dramatic failures and long-term drift. We present a novel background subtraction algorithm designed for pan-tilt-zoom cameras that overcomes this challenge without the need for explicit image registration. The proposed algorithm automatically trains a discriminative background model, which is global in the sense that it is the same regardless of image location. Our approach first extracts multiple features from across the image and uses principal component analysis for dimensionality reduction. The extracted features are then grouped to form a Bag of Features. A global background model is then learned from the bagged features using Support Vector Machine. The proposed approach is fast and accurate. Having a single global model makes it computationally inexpensive in comparison to traditional pixel-wise models. It outperforms several state-of-the-art algorithms on the CDnet 2014 pan-tilt-zoom and baseline categories and Hopkins155 dataset. In particular, it achieves an F-Measure of 75.41% on the CDnet dataset PTZ category, significantly better than the previously reported best score of 62.07%. These results show that by removing the coupling between detection model and spatial location, we significantly increase the robustness to camera motion.     

Fine registration of 3D point clouds fusing structural and photometric information using an RGB-D camera

We address the problem of accurate and efficient alignment of 3D point clouds captured by an RGB-D (Kinect-style) camera from different viewpoints. While the Iterative Closest Point (ICP) algorithm has been widely used for dense point cloud matching, it is limited in its ability to produce accurate results in challenging scenarios involving objects that lack structural features and have significant camera view changes. In this paper, we introduce a new cost function with dynamic weights for the ICP algorithm to tackle this problem. It balances the significance of structural and photometric features with dynamically adjusted weights to improve the error minimization process. Our algorithm also includes a novel outlier rejection method, which adopts adaptive thresholding at each ICP iteration, using both the structural information of the object and the spatial distances of sparse SIFT feature pairs. The effectiveness of our proposed approach is demonstrated by experimental results from various challenging scenarios. We obtained superior registration accuracy than related previous methods, at the same time maintaining low computational requirements.     

基于机器视觉的三维重建技术研究

研究了基于机器视觉的三维重建技术。利用普通的数码摄像机拍摄图片,通过摄像机定标、特征点检测和匹配、基础矩阵和本质矩阵计算来实现图像的三维重建。采用张正友标定方法的相机标定工具箱实现了相机的标定,利用尺度不变特征变换(SIFT)特征点的检测和匹配方法进行了图像特征点的检测和匹配,采用RANSAC算法计算基础矩阵,最后利用相机内参数和由基础矩阵获得的本质矩阵重建物体的特征点,并进行纹理贴图。实验结果表明利用这些图像可以进行物体重建,并且能够很好地反映出物体的三维特征。     

从单幅图像恢复立体景象的新方法

在研究单幅平面图像内在特性的基础上,提出了一种恢复立体视觉景象建模的新方法。对图像进行智能识别处理,可以求得许多线段的特征参数,并由此计算出消隐点和消隐线,从而可自动获得场景的立体结构信息。本算法的特点在于用一个代数表达式统一了三种典型的度量方法,无需传统的相机内校正参数,直接可计算出建模用立体信息。建模结果用VRML格式保存、输出,以便于网上浏览。众多的图像验证了该方法的有效性、适用性。     

Motion and structure from feature correspondences: a review

We present a review of algorithms and their performance for determining three-dimensional (3D) motion and structure of rigid objects when their corresponding features are known at different times or are viewed by different cameras. Three categories of problems are considered, depending upon whether the features are two (2D) or three-dimensional (3D) and the type of correspondence: a) 3D to 3D (i.e., locations of corresponding features in 3D space are known at two different times), b) 2D to 3D (i.e., locations of features in 3D space and their projection on the camera plane are known, and c) 2D to 2D (i.e., projections of features on the camera plane are known at two different times). Features considered include points, straight lines, curved lines, and corners. Emphasis is on problem formulation, efficient algorithms for solution, existence and uniqueness of solutions, and sensitivity of solutions to noise in the observed data. Algorithms described have been used in a variety of applications. Some of these are: a) positioning and navigating 3D objects in a 3D world, b) camera calibration, i.e., determining location and orientation of a camera by observing 3D features whose location is known, c) estimating motion and structure of moving objects relative to a camera. We mention some of the mathematical techniques borrowed from algebraic geometry, projective geometry, and homotopy theory that are required to solve these problems, list unsolved problems, and give some directions for future research