基于机器视觉的图像高维特征智能提取算法  

Intelligent extraction algorithm of image high⁃dimensional features based on machine vision

作　　者：李明磊 赵俊杰 李翔 LI Minglei;ZHAO Junjie;LI Xiang(Citic Daika Co.,Ltd.,Qinhuangdao 066011,China;Guangzhou Gravitational Wave Information Technology Co.,Ltd.,Guangzhou 510000,China)

机构地区：[1]中信戴卡股份有限公司,河北秦皇岛066011 [2]广州引力波信息科技有限公司,广东广州510000

出　　处：《电子设计工程》2023年第18期164-167,173,共5页Electronic Design Engineering

摘　　要：为对图像高维特征完成准确提取,缩小图像的投影向量的偏差量,提出基于机器视觉的图像高维特征智能提取算法。建立视觉坐标系,通过标定图像参数,求解线性摄像机模型表达式,实现基于机器视觉的图像特征匹配。按照尺度空间金字塔构建原则,确定高维图像特征的收敛特性,再联合复杂度参量指标,完成对图像高维特征的提取与智能化处理。实验结果表明,以3D轮毂模型为例,投影向量与真实向量之间的长度差值始终小于0.03 m,与改进SIFT提取方法相比,该方法要符合实际应用要求。In order to accurately extract the high⁃dimensional features of the image and reduce the deviation of the projection vector of the image,an intelligent image high⁃dimensional feature extraction algorithm based on machine vision is proposed.The visual coordinate system is established.By calibrating the image parameters and solving the linear camera model expression,the image feature matching based on machine vision is realized.According to the construction principle of scale space pyramid,the convergence characteristics of high⁃dimensional image features are determined,and then combined with the complexity parameter index to complete the extraction and intelligent processing of high⁃dimensional image features.The experimental results show that taking the 3D wheel hub model as an example,the length difference between the projection vector and the real vector is always less than 0.03 m.Compared with the improved SIFT extraction method,it shows that this method meets the requirements of practical application better.

关 键 词：机器视觉 图像高维特征 智能提取 摄像机模型 空间金字塔 收敛特性 

分 类 号：TN02[电子电信—物理电子学]