面向大规模真正射影像生产的快速遮蔽检测方法

A Fast Approach to Detect Occlusions for Large-scale True Orthophoto Generation

摘要:

为了大规模的真正射影像生产应用，近十几年来，国内外学者努力尝试寻找一种高效、稳健、精确的遮蔽检测方法。基于Z-Buffer技术的遮蔽检测方法已经被证明存在诸多难以克服的问题。本文采用径向投影高度角的变化情况进行遮蔽区域分析，并利用严密的几何关系推导算法的计算过程和范围；然后，利用计算数学理论推导各个步骤的计算复杂度；最后利用微积分的极值理论求出具有最小复杂度的方法。实验表明，基于径向投影高度角变化的方法可以有效解决因地形起伏、分辨率差异和高窄墙体引起的各类伪可见和伪遮挡问题，最优分割的径向扫描方法具有最高的计算效率和最平缓的效率曲线，有利于大规模高质量真正射影像生产。

Abstract:

Previous efforts did not provide an accurate and fast way for generating true othorphoto. This paper tries to solve the problem by analyzing the complexity of algorithm processes and finding the optimum manner to allocate them. In this paper, an optimum segmentation number (OSN) for radial sweep is presented to achieve minimum complexity. First, the scan area, number of azimuth lines and visibility judgment area of radial sweep and spiral sweep method have been discussed with rigorous geometric theory, and then the algorithm complexity of both methods are estimated with mathematical computation theory. At last, minimum complexity of methods is obtained based on extreme point theory of differential calculus. Experiments have demonstrated that the proposed method has the best efficiency, and is efficient to avoid “M-potion” problem, and false occlusions and false visibilities caused by the rolling area, the incompatibility between the DSM and ground image resolution. The deduction and experiments indicate that the proposed method is a robust, accurate, fast and effective approach to generate high quality true orthophoto at a large scale.