随着数字地球与智慧城市等概念的提出与发展，人们对地理空间信息数据的需求由传统的二维数据模型逐步过渡到三维数据模型，无人机倾斜摄影测量是一种常用的采集空间三维数据的技术手段，但由于飞行作业模式的限制，近地区域的数据信息无法获取完整，仅依靠影像数据所建立的三维模型存在着模型扭曲、变形、空洞等问题。因此，针对单一无人机影像数据建模技术的不足，本文研究一种融合倾斜摄影测量多视影像数据和三维激光扫描点云数据的实景三维建模技术，弥补单一技术构建模型的局限性，建立精细的三维实景模型。主要研究内容如下：

(1)无人机在工作中易受到光照的影响，使得影像存在偏亮或偏暗问题，影响影像密集匹配过程中特征提取数量和特征点定位精度。针对上述问题，本文采用了一种结合加权分布的自适应伽马校正(Adaptive Gamma Correction Weighting Distribution, AGCWD)的单参数同态滤波算法对影像进行增强，提高传统的同态滤波对高光区和阴影区的适用性，再使用运动恢复结构(Structure From Motion, SFM)算法生成影像密集匹配点云，实验结果表明经过本文增强后生成的密集匹配点云比传统同态滤波生成的密集点云更加稠密。

(2)由于影像密集匹配点云和三维激光点云是从不同传感器获得的数据，其中存在着尺度不统一问题，本文采用了一种跨尺度点云配准算法，先提取出建筑物点云，使用旋转图像来表示点云的三维点局部特征并计算旋转图像的关键尺度进而求得跨源点云间的尺度差。使用基于快速点特征直方图(Fast Point Feature Histograms, FPFH)的采样一致性初始配准(Sample Consensus Initial Alignment, SAC-IA)算法进行点云粗配准，在点云精配准时，针对传统ICP算法迭代效率低下的问题，引入安德森加速算法来加快迭代过程中收敛速度的速率。通过实验论证本文算法在时间效率和配准精度上都有所提高。

(3)将融合好的点云模型通过Delaunay三角剖分算法建立不规则三角网，通过纹理映射后得到实景三维模型。分别从单一影像数据生成的三维模型和融合影像与点云数据生成三维模型中，选取几处纹理特征明显的区域进行精度评定，结果表明融合影像与点云数据生成的三维实景模型比单一影像数据生成的三维模型更加完整，纹理更加精细。

The inertial navigation system (INS) can realize the navigation and positioning of the carrier only by relying on the measurement information of the inertial components, and has strong autonomy and anti-interference ability. The fiber optic gyroscope (FOG) has become the main inertial component of INS due to its light weight, high reliability, and fast sampling frequency. The output signal of FOG contains various random noises, which need to be filtered and weakened, thereby improving the accuracy of the data. Simultaneously, the INS has its own drift and digital error, which makes the attitude angle and velocity obtained by the integral operation drift over time, especially the position divergence obtained by the quadratic integration. Through the fusion of inertial/visual data, an inertial vision integrated navigation system can be constructed, which can make up for the lack of navigation and positioning accuracy of the inertial navigation system and improve the accuracy of the inertial navigation unit. This paper studies the filtering algorithm of the output signal of the FOG, and discusses the data fusion algorithm of inertial/visual integrated navigation system on this basis. The specific research content of this article is as follows:

1. This article describes the conversion relationship between the coordinate systems of the strapdowm inertial navigation system, the method of calculating the attitude of the carrier, and the update equation and error equation of the inertial navigation. The Allan variance is used to quantify and separate each random noise of the gyroscope, and analyze the signal noise characteristics of the fiber optic gyroscope. The output data of the FOG is analyzed in the time domain, and the random error model of the gyroscope based on Allan variance analysis is established.

2. In order to eliminate the random error of the FOG signal and restore the real signal to the greatest extent, a filtering algorithm based on EMD-SVD fiber optic gyroscope is adopted. Decompose the signal of FOG into multiple intrinsic mode functions (IMF) through empirical mode decomposition (EMD). Using continuous mean square error (CMSE) and Mahalanobis distance (MD) correlation theory to calculate two index parameters, the output signal of fiber optic gyroscope is divided into noise IMFs, mixed IMFs and information IMFs. And the signal reconstruction scheme is designed. The EMD-SVD filtering algorithm is used to filter the mixed IMFs one by one, and the signal is reconstructed to obtain the filtered signal. The filtering algorithm proposed in this paper can effectively suppress the drift of the gyroscope signal and improve the accuracy of the gyroscope's calculation.

3. Based on the fiber optic gyroscope filter algorithm, the advantages and disadvantages of inertial navigation and visual navigation are analyzed. Based on the multi-rate kalman filter (MKF) fusion inertial/visual integrated navigation, the state model and measurement model of the combined system are established for the problem of inconsistent sampling periods when the dual systems are fused, and simulation experiments are carried out. Simulation experiments show that this method has better convergence effect and higher navigation accuracy.