【作者】 李赣华；

【导师】 刘云辉；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2006， 博士

【摘要】 基于摄像机和激光测距仪数据采集方式的三维建模技术一直被研究人员所关注,并在移动机器人的环境感知系统、工业制造、虚拟现实和虚拟增强等众多领域广泛应用。本文系统的分析了不同测量方法的三维建模技术,选择了环境要求低、价格适中的光斑不可见二维激光测距仪和摄像机构建了硬件系统,并建立了相应的传感器控制和数据采集处理软件系统。通过该系统采集数据,对图像特征提取技术、图像匹配技术、系统标定技术和彩色三维建模技术分别进行了方法研究及实际应用。本文目的是:在理论上对基于二维激光测距仪和摄像机的三维建模关键技术进行研究;在实际应用上利用相应技术构建一个三维信息采集重建系统,对室内景物进行彩色三维建模。本文工作主要创新之处如下:1.提出了一个基于模糊阈值的直线提取算法。该算法可以有效解决复杂背景下有畸变图像中的直线提取问题。在铁路沿线的柱状体检测和其距铁轨距离估计的应用中验证了该算法的有效性。2.提出了一个基于Delaunay三角网格的角点匹配算法。该算法主要利用Delaunay三角网格的唯一性,以及相同景物角点对应网格的相似性建立角点匹配。通过对实际图像的实验,证明了该方法对尺度变化和一定程度的旋转和平移变化图象有效。3.对二维激光测距仪标定基本特征进行了分析,提出并设计了基于线特征直角三角形外形的标定物。通过应用于标定二维激光测距仪和摄像机,以及标定二维激光测距仪和旋转云台的实验,证明了设计分析的正确性。4.提出了光斑不可见二维激光测距仪和摄像机的标定算法。有效解决测距数据和图像对应困难的问题。通过与现有算法的仿真和实际采集数据实验比较,证明该方法可以更精确标定设备和更有效融合两种传感器数据。5.提出了二维激光测距仪和旋转云台标定算法。利用标定物正交边测距数据的共线性、正交性和共面性建立约束方程,有效解决了不同转角下测距数据对应困难的问题,有效标定设备。6.提出了异步数据融合的彩色建模算法。解决实际工程应用中高频测距数据和高分辨率图像数据不能同步采集融合的问题。有效融合不同传感器信息,建立了真实感强的室内景物彩色三维点云模型。更多还原

【Abstract】 3D construction technology based on the information collection by camera and laser range finder has attracted great attention all along. And this technology has been widely used in many fields such as the environment sensing using mobile robots, industry manufacture, virtual reality, and augmented reality. By detailed analysis of 3D construction technology based on the different measure method, we chose the 2D laser range finder with invisible stripe and the camera to establish the hardware system, because they have more suitable price and higher adaptability for the environment than other sensors of the 3D measurement. We had also established the software system to control the sensors, to obtain the information, and process the information. Using the measurement of the system, we focused on the methods study and practical application of the image feature extraction, the image matching, the system calibration and the color 3D modeling respectively.Using the camera and the 2D laser range finder, our object is to study the key technologies of 3D model construction in theory, and to establish a 3D color reconstruction system to construct the 3D color model of the indoor scene in practice.The main creativities of this paper are as follows:1. A line extraction algorithm is presented based on the fuzzy threshold. The algorithm can effectively extract the line from the image with the distortion and the complex background. The effectiveness of the algorithm was proved in the application of the colunm detection and the distance estimation along the railway.2. A corner matching algorithm is presented based on the uniqueness of Delaunay triangulated grid and the similarity of the grids for two matching corner sets in the same scene. Experiments on real images demonstrated the good performance of our method, and it was not affected by scale change, less affected by rotation and translation to some extent.3. Based on the main features analysis for the 2D laser range finder calibration, we designed a right-angled triangular checkerboard using the line feature. The efficiency of the checkerboard was demonstrated in the calibration between the laser range finder and the camera. And the calibration between the 2D laser range finder and the pan-tilt platform also proved the checkerboard’s efficiency.4. An algorithm for the extrinsic parameter calibration of the camera and the 2D laser range finder with invisible stripe is presented which effectively solve the data correspondence problem of different sensors. Through the experiment of the simulated data and the real data captured by our system, it was proved that our method could more precisely calibrate the sensors and more effectively fuse the data than the existed calibration methods.5. For the extrinsic parameter calibration of the pan-tilt platform and the 2D laser range finder with invisible stripe, an algorithm is presented based on the colinearity, coplanarity, and orthogonality of the range information. The experimental result proved that it could effectively solve the correspondence problem of the range information with the different pan-tilt platform’s rotation angle.6. An asynchronous information fusion algorithm is presented for 3D color model construction, which could solve the information fusion desynchronizing problem of high frequent range information and high resolution image in the applications. This algorithm could effectively fuse the information of different sensors and construct vivid 3D color point cloud model of indoor sense.更多还原