人工编码点辅助大场景长序列图像的数字近景工业摄影测量方法与技术

With the rapid development of digital manufacturing technology, people put forward new requirements for the precision measurement technology. Because the digital close-range industrial photogrammetry technology has unique advantages in measurement of large workpiece, It has been widely used in various fields. However, the domestic and foreign research on close-range industry photogrammetry is mainly based on two views and three-dimensional reconstruction in a certain extent. The measurement accuracy and the efficiency has been unable to meet the production, so it desiderates advanced measurement method and system with high precision. Therefore, this project proposed theory of close-range industry photogrammetry method of long image sequences of a large-scale scene. Through the analysis of linear geometric constraints between multiple views, this project researched a method of the multifocal tensors solving and computing by the auxiliary of high environmental adaptation ability of artificial encoded point, feature extraction and high precision stereo matching technique using multifocal tensors, created a new principle and a new method of 3D dense points cloud acquisition based on multifocal tensor with high precision and high efficiency; used the bundle adjustment method based on maximum likelihood estimation and cost function with second-order polynomial, researched the global optimization method of long sequence images. This project realized breakthroughs in key technologies and innovation of source in the territory of close-range photogrammetry and 3D reconstruction, provided new ideas and technical support for large workpiece and scene measurement and detection in our country.

随着数字制造技术的迅猛发展,对精密测量技术提出了新的要求，由于数字近景工业摄影测量技术在大型工件测量方面具有独特的优势，在各领域得到广泛应用。然而，国内外对近景工业摄影测量技术的研究主要是围绕着基于两视图的、在某一程度上的三维重建，其测量精度和效率已经不能适应生产，亟需先进的高精度测量方法和系统。为此，本项目提出大场景长序列图像的近景工业摄影测量方法。通过分析多视图之间的线性几何约束关系,研究高环境适应能力的人工编码点辅助多焦点张量求解和计算方法,利用多焦点张量进行特征提取和高精度的立体匹配新技术，创建高精度、高效率的基于多焦点张量的三维密集点云获取新原理和新方法；采用基于极大似然估计和二阶多项式代价函数的集束调整方法，研究长序列图像的整体优化方法。在近景摄影测量和三维重建领域实现源头创新和关键技术突破，为我国大型工件及场景测量与检测提供新的思路和技术支持。

随着数字制造技术的迅猛发展,对精密测量技术提出了新的要求，由于数字近景工业摄影测量技术在大型工件测量方面具有独特的优势，在各领域得到广泛应用。然而，国内外对近景工业摄影测量技术的研究主要是围绕着基于两视图的、在某一程度上的三维重建，其测量精度和效率已经不能适应生产，亟需先进的高精度测量方法和系统。本项目提出了大场景长序列图像的近景工业摄影测量方法。通过分析多视图之间的线性几何约束关系,研究高环境适应能力的人工编码点辅助多焦点张量求解和计算方法,利用多焦点张量进行特征提取和高精度的立体匹配新技术，提出了高精度、高效率的基于多焦点张量的三维密集点云获取新方法；采用基于极大似然估计和二阶多项式代价函数的集束调整方法，解决了长序列图像的整体优化。本项目在近景摄影测量和三维重建领域实现的关键技术，为我国大型工件及场景测量与检测提供新的思路和技术支持。