收稿日期:2022-09-05作者简介:伏子帅(1998-),男,江苏宿迁人,硕士研究生,研究方向为视觉定位、机器人规划及移动机器人导航。基于非线性最小二乘法的二维码视觉定位算法研究伏子帅,钱东海,孙嘉俊,李浩(上海大学机电工程与自动化学院,上海200444)摘要:针对AGV视觉定位算法中存在附加约束的问题,将附加约束的相机坐标系相对于全局坐标系位姿求解问题归纳为非线性最小二乘问题,并采用高斯-牛顿法进行求解。论文所研究的AGV定位算法采用二维码作为路标进行视觉定位,算法首先通过识别经过图像预处理后的二维码,得到该二维码所包含的相对于全局坐标系中的位姿信息;其次建立包含附加约束条件的非线性最优化问题模型,接着通过高斯-牛顿法求出满足包括附加约束条件在内的约束优化问题最优解;最后通过实验验证算法能够解决存在附加约束条件的视觉定位问题,验证了算法的可行性。关键词:二维码;视觉定位;附加约束;非线性最小二乘优化;高斯-牛顿法中图分类号:TP242.6+2文献标识码:ADOI编码:10.14016/j.cnki.1001-9227.2023.01.001ResearchonQRcodevisualpositioningalgorithmbasedonnonlinearleastsquaresmethodFUZishuai,QIANDonghai,SUNJiajun,LIHao(SchoolofMechanicalandElectricalEngineeringandAutomation,ShanghaiUniversity,Shanghai200444,China)Abstract:ToaddresstheproblemofadditionalconstraintsintheAGVvisualpositioningalgorithm,theproblemofsol-vingthecameracoordinatesystemwithadditionalconstraintsrelativetotheglobalcoordinatesystempositionalsolutionisre-ducedtoanonline-arleastsquaresproblemandsolvedusingtheGauss-Newtonmethod.Inthispaper,theAGVpositioningalgorithmusesQRcodesasaroadsignforvisualpositioning.Firstly,thealgorithmidentifiestheQRcodeafterimageprepro-cessingandobtainsthepositionalinformationcontainedintheQRcoderelativetotheglobalcoordinatesystem.Secondly,itestablishesanonlinearoptimizationproblemmodelincludingadditionalconstraintsandthensolvesthisproblemwithaddition-alconstraintsbytheGauss-Newtonmethod.Finally,itisverifiedbyexperimentsthatthealgorithminthispapercansolvetheproblemswithaddi-tionalconstraintsandthefeasibilityofthealgorithmisverified.Keywords:QRcode;visualpositioning;additionalconstraints;nonlinearleastsquaresoptimization;Gauss-Newtonmethod0...
