并行计算在动态摄影测量边缘提取算法中应用
|
摘要
为满足动态摄影测量速度需求,设计一种将Hyper-Q技术应用于双站位相机图像Canny边缘提取算法中的实现方案。通过两个流对采集到的两幅图像分别处理,充分利用GPU计算资源,实现高效并行计算。对300个特征点3种不同分辨率图像进行特征点的Canny边缘检测,实验结果表明,在同样分辨率图像下,基于CUDA的边缘检测算法计算比串行计算算法速度提高了8.8倍,应用Hyper-Q技术后的CUDA程序比串行计算速度提高了11.6倍,图像处理速度显著提高,为双相机动态摄影测量系统在分辨率为4288×2848下实现3Hz测量速度提供思路。
To meet the requirement of operating speed in dynamic photogrammetry,a method of CUDA and Hyper-Q technology was applied to Canny edge extraction of the dual-station camera images.Two image collected through two streams were processed separately,which made full use of GPU computing resources to realize the parallel computing.Canny edge extraction of300 featured points in three different resolution images was implemented in experiment,the results show that the speed of edge detection algorithm based on CUDA is 8.8 times higher than the serial computation accessing the same image resolution,furthermore,the program applied Hyper-Q is 11.6 times higher compared with the serial computation,and so the image processing speed is significantly improved,which provides method for achieving 3 Hz measurement speed under the resolution of 4288×2848 using the dynamic photogrammetry system with dual camera.
To meet the requirement of operating speed in dynamic photogrammetry,a method of CUDA and Hyper-Q technology was applied to Canny edge extraction of the dual-station camera images.Two image collected through two streams were processed separately,which made full use of GPU computing resources to realize the parallel computing.Canny edge extraction of300 featured points in three different resolution images was implemented in experiment,the results show that the speed of edge detection algorithm based on CUDA is 8.8 times higher than the serial computation accessing the same image resolution,furthermore,the program applied Hyper-Q is 11.6 times higher compared with the serial computation,and so the image processing speed is significantly improved,which provides method for achieving 3 Hz measurement speed under the resolution of 4288×2848 using the dynamic photogrammetry system with dual camera.
引文
[1]Teodoro G,Kurc TM,Pan T,et al.Accelerating large scale image analyses on parallel, CPU-GPU equipped systems[C]//Parallel&Distributed Processing Symposium,2012:1093-1104.
[2]WANG Kai,LIU Minxuan,AI Haibin,et al.Fast image matching based on GPU[J].Science of Surveying and Mapping,2014,39(2):129-132(in Chinese).[王恺,刘民选,艾海滨,等.基于GPU的快速影像匹配[J].测绘科学,2014,39(2):129-132.]
[3]LIU Peng,HE Wen,XIAO Weiwei.Rapidly extraction of digital surface model based on GPGPU[J].Journal of Geomatics,2014,39(2):81-84(in Chinese).[刘鹏,何雯,肖巍巍.基于GPGPU技术快速提取数字表面模型[J].测绘地理信息,2014,39(2):81-84.]
[4]XU Xiaochen,DONG Mingli,WANG Jun,et al.A fast target center location algorithm for dynamic vision measurement based on CUDA[J].Computer Engineering&Science,2014,36(12):2378-2385(in Chinese).[许晓臣,董明利,王君,等.基于CUDA的动态视觉测量像面特征点中心快速定位算法[J].计算机工程与科学,2014,36(12):2378-2385.]
[5]Hansch R,Drude I,Hellwich O.Modern methods of bundle adjustment on the GPU[J].ISPRS Annals of Photogrammetry Remote Sensing&Spatial Informa,2016,III-3:43-50.
[6]TANG Bin,LONG Wen.Fast CANNY algorithm based on GPU+CPU[J].Chinese Journal of Liquid Crystals and Displays,2016,31(7):714-720(in Chinese).[唐斌,龙文.基于GPU+CPU的CANNY算子快速实现[J].液晶与显示,2016,31(7):714-720.]
[7]Wilt N.The CUDA handbook:A comprehensive guide to GPU programming[J].CUDA Handbook a Comprehensive Guide to GPU Programming,2013,44(6):147-153.
[8]Lee C,Ro WW,Gaudiot J,et al.Boosting CUDA applications with CPU—GPU hybrid computing[J].International Journal of Parallel Programming,2014,42(2):384-404.
[9]Li H,Yu D,Kumar A,et al.Performance modeling in CUDA streams-a means for high-throughput data processing[C]//IEEE International Conference on Big Data,2015:301.
[10]Luley RS,Qiu Q.Effective utilization of CUDA Hyper-Q for improved power and performance efficiency[C]//IEEE International on Parallel and Distributed Processing Symposium Workshops,2016:1160-1169.
[11]Wang K,Zhang K,Yuan Y,et al.Concurrent analytical query processing with GPUs[J].Proceedings of the Vldb Endowment,2014,7(11):1011-1022.
[12]Cheng J,Grossman M, Mckercher T,et al.Professional CUDA C programming[M]. Washington:Wiley,2014:234-235.
[2]WANG Kai,LIU Minxuan,AI Haibin,et al.Fast image matching based on GPU[J].Science of Surveying and Mapping,2014,39(2):129-132(in Chinese).[王恺,刘民选,艾海滨,等.基于GPU的快速影像匹配[J].测绘科学,2014,39(2):129-132.]
[3]LIU Peng,HE Wen,XIAO Weiwei.Rapidly extraction of digital surface model based on GPGPU[J].Journal of Geomatics,2014,39(2):81-84(in Chinese).[刘鹏,何雯,肖巍巍.基于GPGPU技术快速提取数字表面模型[J].测绘地理信息,2014,39(2):81-84.]
[4]XU Xiaochen,DONG Mingli,WANG Jun,et al.A fast target center location algorithm for dynamic vision measurement based on CUDA[J].Computer Engineering&Science,2014,36(12):2378-2385(in Chinese).[许晓臣,董明利,王君,等.基于CUDA的动态视觉测量像面特征点中心快速定位算法[J].计算机工程与科学,2014,36(12):2378-2385.]
[5]Hansch R,Drude I,Hellwich O.Modern methods of bundle adjustment on the GPU[J].ISPRS Annals of Photogrammetry Remote Sensing&Spatial Informa,2016,III-3:43-50.
[6]TANG Bin,LONG Wen.Fast CANNY algorithm based on GPU+CPU[J].Chinese Journal of Liquid Crystals and Displays,2016,31(7):714-720(in Chinese).[唐斌,龙文.基于GPU+CPU的CANNY算子快速实现[J].液晶与显示,2016,31(7):714-720.]
[7]Wilt N.The CUDA handbook:A comprehensive guide to GPU programming[J].CUDA Handbook a Comprehensive Guide to GPU Programming,2013,44(6):147-153.
[8]Lee C,Ro WW,Gaudiot J,et al.Boosting CUDA applications with CPU—GPU hybrid computing[J].International Journal of Parallel Programming,2014,42(2):384-404.
[9]Li H,Yu D,Kumar A,et al.Performance modeling in CUDA streams-a means for high-throughput data processing[C]//IEEE International Conference on Big Data,2015:301.
[10]Luley RS,Qiu Q.Effective utilization of CUDA Hyper-Q for improved power and performance efficiency[C]//IEEE International on Parallel and Distributed Processing Symposium Workshops,2016:1160-1169.
[11]Wang K,Zhang K,Yuan Y,et al.Concurrent analytical query processing with GPUs[J].Proceedings of the Vldb Endowment,2014,7(11):1011-1022.
[12]Cheng J,Grossman M, Mckercher T,et al.Professional CUDA C programming[M]. Washington:Wiley,2014:234-235.