The special features of Stereo Imaging for LiDAR and hyperspectral sensor are multi-source data and complex algorithm, which will bring huge challenges to embedded real-time processing. To improve system performance, efficient software design is important. In this paper, based on the hardware platform with FPGA+2C6678, a hierarchical parallel model for software design is studied. In intermediate layer, an adaptive dynamic scheduling strategy and a twostage pipeline parallel architecture based on message transmission are presented, which provide efficient connection between the top application design and the bottom hardware environment. The results indicate that this model is strongly supportive for the high-performance of embedded system, and is beneficial for the open and universal design.
|