中文摘要：

提出了一种基于多图形处理器（graphic processing unit,GPU）设计思想的Harris角点检测并行算法,使用众多线程将计算中耗时的影像高斯卷积平滑滤波部分改造成单指令多线程（single instruction multi-ple thread,SIMT）模式,并采用GPU中共享存储器、常数存储器和锁页内存机制在统一计算设备架构（com-pute unified device archetecture,CUDA）上完成影像角点检测的全过程。实验结果表明,基于多GPU的Har-ris角点检测并行算法比CPU上的串行算法可获得最高达60倍的加速比,其执行效率明显提高,对于大规模数据处理呈现出良好的实时处理能力。

英文摘要：

Parallel algorithm of Harris corner detection based on the core concept of Multi- GPU is proposed, so that time-consuming Gaussian image convolution filtering part during the whole image corner detection process can be implemented by many parallel threads. Fi- nally, implementation of this SIMT parallel algorithm using GPU mechanism of shared mem- ory and constant memory and pinned host memory in CUDA is detailed. The Experiments show that the parallel algorithm of Harris corner detection based on Multi-GPU demon- strates substantial improvement up to 60 times speedup than the serial algorithm running in the CPU, is with high efficiency compared with CPU counterpart algorithm, and show great potential for large-scale data processing in real-time processing.