景象匹配算法在多DSP系统中的并行实现  

Parallel implementation of radar scene image match algorithm on multi- DSP system

作　　者：王皎[1] 张天序[1] 颜露新[1] 杨卫东[1] 

机构地区：[1]华中科技大学图像识别与人工智能研究所多谱信息处理技术国防科技重点实验室,湖北武汉430074

出　　处：《微计算机信息》2007年第17期151-153,共3页Control & Automation

基　　金：国家自然科学基金重点项目(60135020);国家重点预研项目(51301050201)

摘　　要：灰度互相关匹配是一种有效的景象匹配算法,其具有适应性强和匹配精度高的特点。但互相关匹配的计算量很大,在实时性要求很高的精确末制导中,匹配速度往往很难满足要求。基于空间并行方法实现了互相关匹配算法的并行处理,通过不同数量处理器的并行处理实验,结果表明:空间并行方法具有很高的加速比和并行效率,是一种适合互相关匹配算法的有效的并行处理方法。当图像尺寸增大计算量增加时,可以采取增加处理器提高系统处理能力的办法使算法依然满足实时性要求。　Gray intensity based correlation method is an effective image matching algorithm, which has better adaptability and high matching precision. But it does not meet real- time guidance application demand for its complicated and intensive computation. To achieve the performance of real- time processing an implementation of correlation algorithm is proposed which bases on space parallelism. The experiment of parallel implementation on different number of processors shows that the method of space parallelism can achieve high speedup and efficiency for gray intensity correlation matching algorithm. To attain the same performance more processors could be added to gain much powerful computing ability when the image size and the computation increased.

关 键 词：景象匹配 灰度互相关 并行处理 DSP 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]