结合视觉惯性模组的室内三维布局鲁棒重建方法

针对使用传统单目相机的全自动三维重建方法结果精确度差和整体结构理解缺失等问题,提出一种结合视觉惯性里程计和由运动到结构的全自动室内三维布局重建系统.首先利用视觉里程计获得关键帧图像序列和对应空间位置姿态,并利用运动恢复结构算法计算精确相机位姿;然后利用多图视立体几何算法生成高质量稠密点云;最后基于曼哈顿世界假设,针对典型的现代建筑室内场景,设计一种基于规则的自底向上的布局重建方法,得到最终房间外轮廓布局.使用浙江大学CAD&CG实验室场景现场扫描数据集和人工合成的稠密点云数据集作为实验数据,在Ubuntu 16.04和PCL 1.9环境下进行实验.结果表明,文中方法对三维点云噪声容忍度高,能够有效地重建出室内场景的三维外轮廓布局.     

利用真实场景中的角结构约束进行相机自定标

首先讨论了真实场景中的角结构几何约束，然后利用这种几何约束提出了一种新的相机自定标算法，利用单幅透视投影图像中的角结构计算出相机的焦距、平移向量和旋转矩阵的初始值，然后利用两幅图像中的几何结构对相机内外参数进行优化，由于在求取相机参数初始值的时候只用到了一幅图像，这样就避免了在相机定标过程中可能出现的临界运动序列（critical motions sequence)，从而避免了临界运动序列引起的相机定标退化问题，提高了相机定标过程的鲁棒性，用两幅图像中的结构约束对初始值进行优化，进一步提高了结果的精确度，实验结果证明算法是强壮的。     

基于改进SFM的三维重建算法研究

针对现有运动恢复结构算法重建模型存在点云稀疏等问题,提出一种利用不同匹配数据进行模型重建的算法。首先通过对比上下文直方图(CCH)生成匹配数据,利用M估计抽样一致(MSAC)估算图像基础矩阵,进而分解得到平移和旋转矩阵,并根据相机内参计算投影矩阵,然后利用KLT匹配算法更新匹配数据,最后三角化生成三维点云。该算法匹配精度高,图像基础矩阵易于收敛,通过位移实现特征点匹配,弥补了图像低频区域匹配数据不足的缺陷。实验结果表明,与现有算法相比,该算法生成的点云更致密;在真实环境下,该算法可用于物体三维重建。     

一种基于4对图像对应点的欧氏重建方法

摄像机自标定算法通常是非线性的，为了得到线性的方法，提出了一种在RANSAC框架下由4对图像对应点线性标定摄像机并对场景进行鲁棒性欧氏重建的方法。当摄像机作两组平移运动时，若在两组平移运动之间摄像机具有不同的姿态，则从4对图像对应点可以线性地重建场景的欧氏几何。模拟实验和真实图像实验均证明了本文方法的可行性。     

基于未标定图像的三维重建算法

提出一种基于多幅未标定图像的三维重建算法。在标记点匹配的基础上进行射影重建,通过施加度量约束将射影重建升级为欧氏重建,即利用未标定的透视图像恢复相机的内、外部参数以及标记点的三维空间坐标,实现场景的三维重建。标记点易于进行点对精确匹配,较手动拾取匹配提高了效率。实验结果表明,利用该算法能够大幅减小再投影误差。     

基于一种优化的配准算法的图像超分辨率重建

提出了一种优化的图像配准算法.该算法充分结合了时域基于全局运动模型的配准算法和频域基于傅立叶变换的配准算法,并运用到超分辨重建中.假设序列低分辨率图像之间存在旋转和平移,首先运用全局配准法求出旋转参数,对图像进行旋转补偿,然后对补偿后的图像用频域配准法估计平移参数,针对估计出的旋转参数和平移参数对序列低分辨率图像进行超分辨率图像重建,实验表明该方法能较有效地提高图像分辨的效果,并且本文还阐述了低分辨率图像帧数的影响.     

特种复杂试验装置中模型动态建模系统设计

为克服特种复杂试验装置中模型不可接触、传统动态建模方式周期长等困难,设计一种基于运动恢复结构算法的动态建模系统,实现实时的动态建模,以更直观和精确的指导试验。试验装置内设置车道线,使用智能寻迹车进行轨迹识别,环绕模型行进。智能寻迹车上同时搭载光学相机和红外相机,同视角拍摄特种复杂试验装置中模型的图像。使用运动恢复结构算法,从光学相机拍摄的图像重建模型的三维结构。模型上设置有环状编码标记点,利用标记点的坐标增加重建的精度。使用改进的运动恢复结构算法,从红外相机拍摄的图像重建红外三维结构,并建立起与光学图像生成的三维结构的对应关系。系统最终获得具有温度信息的可视化模型,并可使用头戴式虚拟现实设备进行沉浸式体验。     

基于平面的建筑物表面模型重建算法的研究

针对建筑物模型的规则性，提出了一种基于平面的建筑物模型重建算法，可以从单幅透视图像恢复建筑物的表面模型．该算法主要分为相机定标、基平面的提取、平面位置和方向的计算等几个子过程．相机定标主要用于求解相机的焦距，是一个非常重要的部分．该算法以建筑物场景中的几何结构作为约束条件，从单幅图像中求解相机的焦距；然后计算基平面位置和法向；最后通过交互式操作指明场景中各平面之间的相互关系，递归求解各平面的位置和法向，达到根据图像重建建筑物场景表面模型的目的．     

基于差分演化算法的MR图像平移运动伪影校正

在磁共振(Magnetic resonance,MR)成像过程中,病人的自主运动或生理性运动会使重建的图像产生伪影,严重影响医生的诊断.为了校正MR图像平移运动伪影,本文提出了一种基于差分演化算法的校正方法.首先利用非兴趣区的伪影数据,建立关于K空间偏移相位的约束条件方程组.然后采用差分演化算法搜索最优平移位移量,由此计算出相位偏差量并修正图像.同时提出一个新的价值函数,以此作为演化计算终止的判断条件,实现伪影图像的自动校正.实验结果表明,该算法能够有效地修正含有平移运动伪影的MR图像,且不需要反复的人工掩膜处理.     

基于立体视觉的球形机器人定位方法

针对球形机器人定位问题,提出了基于立体视觉的球形机器人定位方法.通过双目相机采集环境图像序列,提取Shi-Tomasi特征点,计算尺度不变特征变换(SIFT)特征描述符,并利用欧氏距离进行立体匹配;通过KLT算法进行特征点跟踪;采用解析法求解机器人在前后帧图像之间的位姿变化量;同时采用特征点筛选、RANSAC算法和卡尔曼滤波等方法,提高运动估计的准确性和鲁棒性.实验结果验证了所提出方法的可行性.     

Depth recovery and affine reconstruction under camera pure translation

This paper addresses the problems of depth recovery and affine reconstruction from two perspective images, which are generated by an uncalibrated translating camera. Firstly, we develop a new constraint that the homography for the plane, which is orthogonal to the optical axis, is determined only by the epipole and the plane's relative distance to the origin under camera pure translation. The algorithm of depth recovery is based on this new constraint, and it can successfully avoid the step of camera calibration. With the recovered depth, we show that affine reconstruction can be obtained readily. The proposed affine reconstruction does not need any control points, which were used to expand the affine coordinate system in existing method. Therefore, it could avoid the step of non-planarity verification as well as the errors from the control points. Error analysis is also presented to evaluate the uncertainty for the recovered depth value. Finally, we have tested the proposed algorithm with both simulated data and real image data. And the results show that the proposed algorithm is accurate and practical.     

基于4目阵列的计算机视觉三维重建算法

目前计算机视觉三维重建方法因需布置和标定摄像机环形拍摄场或者需要结构光而存在应用局限性问题,且算法不稳定。为此提出一种将摄像机阵列和图像配准有机结合的4目阵列重建算法,该算法不需要结构光和现场标定摄像机。经过基于包含光照和阴影的复杂室内仿真图像的实验表明,该方法能稳定有效地进行密集点云重建,且能克服现有重建方法的应用局限性与不稳定等缺陷。     

Euclidean upgrading from segment lengths: DLT-like algorithm and its variants

In this paper, how to calibrate a fixed multi-camera system and simultaneously achieve a Euclidean reconstruction from a set of segments is addressed. It is well known that only a projective reconstruction could be achieved without any prior information. Here, the known segment lengths are exploited to upgrade the projective reconstruction to a Euclidean reconstruction and simultaneously calibrate the intrinsic and extrinsic camera parameters. At first, a DLT(Direct Linear Transformation)-like algorithm for the Euclidean upgrading from segment lengths is derived in a very simple way. Although the intermediate results in the DLT-like algorithm are essentially equivalent to the quadric of segments (QoS), the DLT-like algorithm is of higher accuracy than the existing linear algorithms derived from the QoS because of a more accurate way to extract the plane at infinity from the intermediate results. Then, to further improve the accuracy of Euclidean upgrading, two weighted DLT-like algorithms are presented by weighting the linear constraint equations in the original DLT-like algorithm. Finally, using the results of these linear algorithms as the initial values, a new weighted nonlinear algorithm for Euclidean upgrading is explored to recover the Euclidean structure more accurately. Extensive experimental results on both the synthetic data and the real image data demonstrate the effectiveness of our proposed algorithms in Euclidean upgrading and multi-camera calibration.     

Some Results on Minimal Euclidean Reconstruction from Four Points

Methods for reconstruction and camera estimation from miminal data are often used to boot-strap robust (RANSAC and LMS) and optimal (bundle adjustment) structure and motion estimates. Minimal methods are known for projective reconstruction from two or more uncalibrated images, and for “5 point” relative orientation and Euclidean reconstruction from two calibrated parameters, but we know of no efficient minimal method for three or more calibrated cameras except the uniqueness proof by Holt and Netravali. We reformulate the problem of Euclidean reconstruction from minimal data of four points in three or more calibrated images, and develop a random rational simulation method to show some new results on this problem. In addition to an alternative proof of the uniqueness of the solutions in general cases, we further show that unknown coplanar configurations are not singular, but the true solution is a double root. The solution from a known coplanar configuration is also generally unique. Some especially symmetric point-camera configurations lead to multiple solutions, but only symmetry of points or the cameras gives a unique solution.     

Catadioptric Stereo Using Planar Mirrors

By using mirror reflections of a scene, stereo images can be captured with a single camera (catadioptric stereo). In addition to simplifying data acquisition single camera stereo provides both geometric and radiometric advantages over traditional two camera stereo. In this paper, we discuss the geometry and calibration of catadioptric stereo with two planar mirrors. In particular, we will show that the relative orientation of a catadioptric stereo rig is restricted to the class of planar motions thus reducing the number of external calibration parameters from 6 to 5. Next we derive the epipolar geometry for catadioptric stereo and show that it has 6 degrees of freedom rather than 7 for traditional stereo. Furthermore, we show how focal length can be recovered from a single catadioptric image solely from a set of stereo correspondences. To test the accuracy of the calibration we present a comparison to Tsai camera calibration and we measure the quality of Euclidean reconstruction. In addition, we will describe a real-time system which demonstrates the viability of stereo with mirrors as an alternative to traditional two camera stereo.     

Euclidean reconstruction from contour matches

This paper presents a method that recovers an Euclidean structure from contour matching results. If three or more views are available, structure reconstruction without camera calibration is possible. This is the first attempt at structure reconstruction from contour matches, though there are works on structure reconstruction from point matches. We also propose a new method that integrates more than one reconstructed 3D structure whose coordinates are represented by different camera projection matrices, so that the structures are uniformly represented by the projection matrix of the first camera.     

计算机视觉测量系统的误差模型分析

基于相机的钻孔模型，研究了相机光学系统产生的4种非线性误差，并在有径向失真和图像中心偏移失真的情况下，用模拟数据对匹配误差和不同量化等级在重建过程中的传播过程和影响作了详细讨论，为视觉测量系统达到高精度的三维重建提供了可靠的误差模型。     

An improved algorithm for two-image camera self-calibration and Euclidean structure recovery using absolute quadric

We present an improved algorithm for two-image camera self-calibration and Euclidean structure recovery, where the effective focal lengths of both cameras are assumed to be the only unknown intrinsic parameters. By using the absolute quadric, it is shown that the effective focal lengths can be computed linearly from two perspective images without imposing scene or motion constraints. Moreover, a quadratic equation derived from the absolute quadric is proposed for solving the parameters of the plane at infinity from two images, which upgrades a projective reconstruction to a Euclidean reconstruction.     

Stratification approach for 3-D euclidean reconstruction of nonrigid objects from uncalibrated image sequences.

This paper addresses the problem of 3-D reconstruction of nonrigid objects from uncalibrated image sequences. Under the assumption of affine camera and that the nonrigid object is composed of a rigid part and a deformation part, we propose a stratification approach to recover the structure of nonrigid objects by first reconstructing the structure in affine space and then upgrading it to the Euclidean space. The novelty and main features of the method lies in several aspects. First, we propose a deformation weight constraint to the problem and prove the invariability between the recovered structure and shape bases under this constraint. The constraint was not observed by previous studies. Second, we propose a constrained power factorization algorithm to recover the deformation structure in affine space. The algorithm overcomes some limitations of a previous singular-value-decomposition-based method. It can even work with missing data in the tracking matrix. Third, we propose to separate the rigid features from the deformation ones in 3-D affine space, which makes the detection more accurate and robust. The stratification matrix is estimated from the rigid features, which may relax the influence of large tracking errors in the deformation part. Extensive experiments on synthetic data and real sequences validate the proposed method and show improvements over existing solutions.