| 摘要: |
| LK(Lukas-Kanade)光流法在运动目标检测和跟踪领域具有广泛应用，但其计算复杂、速度慢，难以适应异构硬件平台。为实现LK光流法在不同平台上的高效运行，设计了一种基于开放式计算语言（OpenCL）的LK光流法并行算法。该算法通过将二维图像上像素点上的稠密计算映射到多线程上实现数据并行，并基于OpenCL平台的共享内存等优化方法减小了主机内存与设备内存数据传输。实验测试表明，该算法相比于多核CPU下的基础OpenCV函数库中的LK算法获得了最高31倍的加速比，同时在速度上与统一计算设备体系结构（CUDA）加速的LK光流法相近。此外，还在多种不同设备下验证了加速算法的平台通用性。 |
| 关键词: 行为识别 目标跟踪 Lukas-Kanade光流法 OpenCL异构计算 GPU并行加速 |
| DOI： |
|
| 基金项目:国家自然科学基金资助项目（61772417，61634004，61602377）；陕西省自然科学基础研究计划项目（2018JM4018）；陕西省科技统筹创新工程项目（2016KTZDGY02-04-02）；陕西省重点研发计划（2017GY-060） |
|
| A Lukas-Kanade optical flow parallel acceleration algorithm based on OpenCL |
| WU Jin,LI Qiaoshen,MIN Yu,MA Simin |
| (School of Electronic Engineering,Xi′an University of Posts and Telecommunications,Xi′an 710121,China) |
| Abstract: |
| Lukas-Kanade(LK) optical flow algorithm has been widely used in the field of moving target detection and tracking,but its computation is complex and slow,so it is difficult to be adaptive to heterogeneous hardware platforms.In order to implement the efficient operation of LK optical flow algorithm on a variety of platforms,a parallel LK optical flow algorithm based on OpenCL is designed.The algorithm implements data parallelism by mapping the dense calculations at the dpi points on the two-dimensional image to multiple threads.It reduces the data transfer between the host memory and the device memory based on the OpenCL platform's shared memory and other optimization methods.The experimental tests show that this algorithm achieves a 31-fold speedup compared with the LK algorithm in the underlying OpenCV library of multi-core CPUs while its calculation speed is similar to that of the LK optical flow algorithm with the acceleration of CUDA.In addition,the platform generality of the accelerated algorithm has been validated on a variety of devices. |
| Key words: behaviour recognition target tracking Lukas-Kanade optical flow algorithm OpenCL heterogeneous computing parallel acceleration with GPU |
《电讯技术》编辑部 