中文摘要：

连接点提取是机载三线阵CCI）影像自动空中三角测量的重要前提。针对机载三线阵CCD影像成像特点以及数据量大、处理耗时等劣势，研究了三线阵影像基于POS引导的GPU加速匹配技术，并结合核线约束及RANSAC方法剔除误匹配点。选取三种不同地物特征影像进行匹配试验，验证了该匹配方法的稳定性，同时通过与传统SIFT匹配方法的对比，验证了基于GPU加速的匹配方法具有较好的加速比，也为进一步开展平差技术研究提供技术支持。

英文摘要：

Tie point extraction is an important prerequisite for automatic aerial triangulation of airborne three-line array CCD images. Taking the imaging features of airborne three line array CCD images and disadvantages like large amount and time consuming into consideration, a POS-guided GPU acceleration matching technology for three-line array image is studied, and the mismatching points are eliminated using epipolar line constraint and RANSAC method. Three groups of stereo image pairs with different typical ground features are selected to test the stability of GPU acceleration matching method. Comparison between the traditional SIFT method and proposed method in this paper proves that the GPU acceleration matching has a better acceleration ratio. The study can also provide support for further aerial triangulation adjustment technology research.