基于FPGA的高效机载视频采集及预处理方法

为提高机载视频编码系统的数据采集及预处理性能,以现场可编程门阵列(FPGA)为硬件平台,研究了多模式机载视频采集、颜色空间转换和视频数据传输的高效处理方法.针对机载应用需要实时采集不同模式视频的特点,设计了一种可靠的视频采集策略,通过引入错误自检机制,可以实时监测视频采集的正确性,避免视频数据的错误积累;针对机载视频颜色空间转换预处理中浮点乘法浪费计算资源和增加系统功耗的问题,设计了一种基于高低位分离的截断式查找表乘法器,减少了存储空间和计算位宽,结合流水线处理技术实现了一种高效视频颜色空间转换方法,在保证计算精度和性能的同时,处理功耗最大降低了27%;针对FPGA处理器与系统核心编码处理器(DSP)之间存在大量视频数据的频繁传输特点,结合SRIO(Serial Rapid I/O)链路的传输方式,设计了一种以FPGA为控制核心的数据交互机制,减轻了DSP的处理负担使其专注于视频编码运算,提高系统性能.

Method for airborne video capture and preprocessing based on FPGA

Video capture, color space conversion and video data transmission were researched in order to improve video capture and pre-processing performance of airborne multi-mode video coding system. Taking into account the requirements of real-time multi-mode video capture, a robust video capture strategy based on field-programmable gate array (FPGA) was designed. It monitors the correctness of video capture on-time to avoid accumulation of errors by adopting self-checking mechanism. To solve the problem of wastes of computing resources and increasement of power consumption in floating-point multiplication, a truncated look-up table multiplier based on the separation of high and low bits was designed to save memory and computation bandwidth. It achieves an efficient color space conversion combining pipeline technique, with power consumption decreasing 27% at most without loss of computing precision and processing performance. Considering the frequent transmission of mass data between FPGA and digital signal processor (DSP), a data interaction mechanism taking FPGA as the control core was proposed, adopting the transmission mode of serial rapid I/O (SRIO). The burden of DSP was alleviated due to concentrating on video coding, and thus the application performance was improved.