基于OpenCL的多标靶加速定位方法  

Accelerated Multi-Target Positioning Method Based on OpenCL

作　　者：王瑾瑜 罗剑波[1] WANG Jinyu;LUO Jianbo(School of In f ormation Science and Engineering,Zhejiang Sci-Tech University,Hangzhou 310018,China)

机构地区：[1]浙江理工大学信息科学与工程学院,浙江杭州310018

出　　处：《软件工程》2024年第8期57-61,共5页Software Engineering

基　　金：浙江省公益技术应用研究资助项目(LGG22E050051)。

摘　　要：针对现有人工标靶定位方法精度不高且效率低的问题,提出一种基于ICP(Iterrative Closest Point)的定位标靶中心算法,并在所提出的定位方法的实现上进行加速,对检测到的二维边缘点集在CPU中建立VP-tree(Vantage Point Tree)数据结构后,传入图形处理器(Graphics Processing Unit,GPU)缓冲区中,使用全称为开放计算语言(Open Computing Language,OpenCL)框架并行计算,并且使下一时刻的边缘点云继承上一时刻的刚性变换矩阵,减少迭代次数并加快收敛速度,实现实时定位。经实验验证,本文算法在1.6 m的视场范围,定位精度约为0.0819 pixel,平均绝对误差约为0.0261 mm。实验结果验证了该方法具有可行性且算法有效。To address the issues of low accuracy and inefficiency in existing manual target positioning methods,this paper proposes a target center positioning algorithm based on the Iterative Closest Point(ICP).To improve the efficiency of the positioning method,the detected 2D edge point set is processed in the CPU to build a Vantage Point Tree(VP-tree)data structure,which is then transferred to the Graphics Processing Unit(GPU)buffer.Parallel computation is performed using the Open Computing Language(OpenCL)framework,allowing the next moment's edge point cloud to inherit the rigid transformation matrix from the previous moment,reducing the number of iterations and accelerating convergence speed to achieve real-time positioning.Experimental validation shows that the proposed algorithm achieves a positioning accuracy of approximately 0.0819 pixels and an average absolute error of about 0.0261 mm within a field view of 1.6 meters.The experimental results verify the feasibility and effectiveness of the proposed method.

关 键 词：标靶定位 OPENCL 迭代最近点 GPU 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]