摘要
针对DSP/BIOS环境下在线视频图像火焰检测算法实现中存在的巨量数据运算和提高CPU利用率之间的矛盾,提出了以TMS320DM642为处理器的视频图像处理硬件系统构建,并在DSP/BIOS下设计了基于任务通信的实时视频图像火焰检测系统。此外,给出了基于颜色和火焰蔓延动态特性的检测规则,利用Ping-pong缓存技术实现了EDMA数据传输和CPU处理的并行操作,大大提高了CPU的利用率。试验结果表明,该算法不仅能有效地提取火焰区域,而且能使单帧图像的处理速度较无缓存技术和缓存技术分别提高60倍和1.55倍。
Aiming at the eontradiction between huge date calculations and improving CPU utilization in flame detection of online video images under the DSP/BIOS, a hardware architecture based on TMS320DM642 processor for video image processing system was proposed. Moreover, a real-time video image flame detection system was further designed based on task communication of RF5 under DSP/BIOS environments. Furthermore, a detection rule was given, based on color and flame dynamic spreading features. The CPU utilization rate was greatly improved by using Ping-pong buffering for implementing parallel operation of EDMA data transfer and CPU image processing. The research resuhs show that the algorithm could not only extract flame regions effectively, but also make the processing speed of a single frame reach 60 times and 1.55 times over conventional method using no cache and cache.
出处
《机电工程》
CAS
2009年第12期34-37,61,共5页
Journal of Mechanical & Electrical Engineering
基金
国家自然科学基金资助项目(60775016)
浙江省重大科技专项资助项目(2007C13062)
