双处理器实时视频拼接系统

Real-time video mosaic systems based on double processors

当一些视频系统需更大观测范围,同时获得完整场景是一个问题.针对这种情况,提出一套基于FPGA和DSP双处理器协同工作的视频拼接系统,设计了包括双处理器、FLASH程序存储器、SDRAM数据存储器、SRAM数据存储器、视频AD和视频解码器等硬件结构.系统采集两路视频信号输入,对其进行数字视频转换,然后由FPGA做基本图像处理,并将其送入SDRAM进行存储.DSP将初步处理后的图像数据取出,并对其进行图像配准和图像融合.为了保证图像的实时性,系统采用实时高效易实现的改进的PSO算法实现图像配准,采用平均加权法实现图像融合.实验结果表明,此系统可对两幅720×576分辨率、25帧/秒的视频图像进行实时拼接,满足所需大视场视频图像要求.

It is difficult for some video systems to capture the whole scene simultaneously. In view of this situation, a video mosaic system based on FPGA and DSP dual-processors was proposed. A hardware architecture including dual-processors, FLASH program memory, SDRAM data memory, SRAM data memory, video AD and video decoder was designed. Two video input signals were captured and converted to digital signals and were stored in the SDRAM after basic image processing by FPGA. The image data readout, registration and fusion were processed by DSP. To ensure the real-time property of the system, the improved particle swarm optimization （IPSO） algorithm was adopted to process the image registration; and the image fusion was processed by the weighted average method. The test results show that two images were aequired and stitched simultaneously at resolution of 720X 576 and 25FPS.