几何约束自检校的无人机高精度影像定向与纠正算法研究

Recent years, a number of types of UAV(Unmanned aerial vehicle) systems with various onboard sensors have been developed for civilian applications such as homeland security, urban and city planning, forestry fire monitoring, quick response measurements for emergency disaster, Earth science research, and humanitarian observations. Recent developments in the vehicles themselves and associated sensing systems make these platforms increasingly attractive to the geoscience community. UAV platforms and imaging and sensing systems that are adaptable to these platforms facilitate unique capabilities in Earth observation for both research and operational monitoring purposes. .Compared with traditional satellite and aerial remote sensing images, UAV image has the features of lower cost, high resolution and quick response. However, its stabilization may be affected dramatically by the light weight of the vehicle, so it's difficult to take the image orientation with high accuracy. Focusing on the features of urban area large-scale UAV image, this project proposes uncalibrated camera based method of high accurate UAV image orientation and geometric rectification. With the theory of generalized point photogrammetry, the project develops IAC (Image of Absolute Conic) based image calibration and orientation algorithm, and also the accruracy anylysis. Furthermore, using the line feature in the facade of the buiding, it addresses an improvement method of high-accuracy Orthorectification Model with Self-Geometric Constraint.

近年来，低空遥感技术受到众多应用领域的高度关注。其中，光学传感器由以往的可量测型专业相机逐渐被非量测相机所替代。和卫星遥感影像、传统航空遥感影像相比，低空无人机光学遥感影像具有低成本、高分辨率、地面特征丰富、应用灵活等特点。但影像的稳定性受飞行器的姿态变化影响较大，难以进行高精度的影像定向。本项目以无人机为数据获取平台，针对城市区域大比例尺低空影像的特点，提出一种基于未标定相机的高精度UAV影像定向及几何纠正方法，该方法在广义点（灭点）理论的支持下，研究基于线特征IAC约束的的影像精确定位、定向算法，以及三步分解法的模型误差传播与精度分析。同时，充分利用低空影像中建筑物目标的线特征，采用基于广义点几何约束的区域网平差算法对几何纠正进行精度改进。通过实验分析，得到低空影像区域网平差的精度分析报告，实现UAV影像定向参数与纠正算法的优化。

近年来,无人飞机UAVs 已受到众多应用领域的高度关注。以无人飞机为平台的航空遥感传感器由以往的可量测型像机逐渐被非量测型摄像机所替代, 其优点是像机价格低廉,可以获取连续的影像数据流。和卫星遥感影像、传统航空遥感影像相比, UAV影像的特点是: 图像精度高, 影像清晰, 地面特征丰富, 但影像的稳定性受飞行器的姿态变化影响较大，难以进行高精度的影像定向。对这种UAV影像进行高精度定向及几何纠正后, 可以获得高分辨率的正射影像, 该影像在灾害监测、资源勘察、紧急救援、国土安全评估及空间决策支持等应用领域具有广阔的应用前景。但UAVs影像的重叠度大,且相机为非专业航空数码相机，不具有传统航空像片的可量测性, 因此, 对该影像进行几何纠正处理是其首要问题。.本文针对载有非量测相机的无人机影像,利用城市地区的人工建筑物特征，研究未标定影像的高精度定向及影像纠正算法。在相机标定方面，充分利用建筑物线特征，提出了基于三轴多像线特征的相机标定方法。该方法无需任何标定物或标志点，只需要建筑表面线特征进行投影几何约束，即可完成非量测相机的标定。与此同时，利用建立的线特征几何约束，将约束条件引入至总体纠正平差算法中，即可完成影像定向模型。试验表明，该方法大大提高了影像定标定向精度。