改进SIFT特征粒子群优化的机器人视觉图像配准模型研究  

Optimization of robot visual images registration model based on SIFT particle swarm improvement

作　　者：李莹[1] 王哲[2] 李毛 

机构地区：[1]陕西工业职业技术学院信息工程学院,咸阳712000 [2]陕西工业职业技术学院数控工程学院,咸阳712000 [3]陕西省机电工程学校电气专业,咸阳712000

出　　处：《现代制造工程》2015年第9期34-37,共4页Modern Manufacturing Engineering

摘　　要：针对机器人视觉图像尺度小、重复率低、数据量大和实时性差等特点,以提高图像配准效率、减少配准时间为目的,提出了改进SIFT特征粒子群优化的机器人视觉图像配准CSIFT-QPSO算法。CSIFT-QPSO算法通过建立S金字塔对SIFT算法进行改进,达到降低空间复杂度和减少极值点数量的目的,并将改进后的SIFT算法与QPSO算法中适应值函数相融合,将当前检测到的极值点与最近更新的目标模块相配准,从而更具鲁棒性地应对机器人视觉图像的配准问题。通过Matlab软件进行仿真,仿真结果表明:CSIFT-QPSO算法具有较高的配准精度和较少的配准时间,适用于对实时性要求较高的机器人视觉图像配准系统。Against the characteristics of robot visual images small-scale, low overlap rate and large amount of data, in order to im- prove the efficiency of image registration, reducing the time for the purpose of registration, proposes the improved SIFT feature particle swarm optimization robot image registration CSIFT-QPSO algorithm. To reduce the muhi-scale spatial and the number of feature points, the algorithm through the establishment of the S-layer pyramid, integrate the fitness function of the improved algo- rithm SIFT and QPSO,to match the currently detected feature points and the target module updated recently, thus more robust to deal with the registration problem of robot visual images. Finally, Matlab simulation results show that the improved algorithm has higher matching accuracy and less matching time,which is suitable for real-time robot visual images registration system.

关 键 词：视觉图像 CSIFT-QPSO算法 尺度空间 适应值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]