中文摘要：

将嵌入式系统与实时摄影测量相结合，提出了一种嵌入式系统上ARM＋DSP架构的机载影像实时拼接方法。ARM＋DSP架构中，由ARM负责数据接收和结果的保存与显示，DSP负责数据处理过程，通过数据缓冲和内存共享机制，实现数据接收和数据处理的并行化以及ARM与DSP数据交互的零开销。同时，使用递推的纠正方式和最邻近采样进一步加快拼接处理过程。实验验证，针对GSD为2．5m、像幅为3027像素×5530像素的Bayer影像数据，生成GSD为5m的具有地理参考的全景图，该方法能保证每秒1张的拼接速度，完全满足航摄曝光的时间间隔需求。

英文摘要：

Combining an embedded system and photogrammetry, this paper proposes a real-time air- borne image mosaic method on an embedded system based on the ARM＋ DSP system architecture. Characterized by its small volume, low power, and high integration density, an embedded system is a very promising solution for real time airborne photogrammetry problems. In our work, ARM receives the image data and DSP processes that data, then ARM saves and displays the result. Data buffering and memory sharing are used between ARM and DSP, so receiving and processing data can be parallelized and exchanging data between ARM and DSP uses little time. Meanwhile, recurrence rectification and nearest neighbor interpolation methods are used to accelerate the mosaic process. Experiments show that the proposed method can generate geo referenced mosaic images in GSD 5 m in 1 FPS （frames per second）, for Bayer images at a size of 3 027×5 530 pixels and GSD 2.5 m. The real-time airborne image mosaic method using the ARM＋DSP architecture meets the exposure interval require ment for aerial photogrammetry.