航天影像间的自动转点方法

Automatic Extraction of Connection Point in Aerospace Images

航天影像间的自动转点是航天影像自动空中三角测量的主要环节，寻找一种鲁棒性好、耗时少的自动转点方法至关重要。本文针对传统人工航天影像转点耗时、耗力的问题，提出了一种全自动的航天影像转点方法，实现了批量航天影像自动化转点，大幅减少了生产作业的成本。该方法首先在读取影像边角点空间坐标的基础上，自动计算影像重叠区域；然后，在利用GPU加速SIFT算法的基础上进行基于分块的航天影像匹配，从而实现了航天影像问的自动转点过程；最后，利用“天绘一号”卫星的三线阵影像进行实验，并利用最小二乘影像匹配等方法对实验结果进行了分析。实验表明，该自动转点方法耗时较少、精度可靠，能够较好地减少作业成本，具有一定的可行性。

It is a key link to find an automatic extraction of connection point in aerial triangulation measurement. In order to solve the time -consuming problem with the traditional methods, this paper proposes a new automatic turning point method, which realizes the batch aerospace images processing and greatly reduces production costs. In this way, firstly the overlapping areas are automatically calculated based on reading the image corner point coordinates. Then block matching is done by using GPU - based SIFT algorithm, thus it realizes the automatic extraction of connection points. Finally this paper conducts the experiment based on three - line array CCD images from Mapping satellite - I and analyzes the results by using the least squares image matching method. The experiments show that the automatic extraction method takes less time and has reliable accuracy. Also it can reduce the operation costs and has certain feasibility.