烟梗按序上料码放位置RANSAC优化识别方法  

RANSAC Optimal Identification Method for Tobacco Stem Stacking Position in Sequence

作　　者：孟瑾[1] 王伟 崔建华 石怀忠 MENG Jin;WANG Wei;CUI Jian-hua;SHI Huai-zhong(Henan Zhongyan Industry Co.,Ltd.Anyang Cigarette Factory,Henan Anyang 450016,China)

机构地区：[1]河南中烟工业有限责任公司安阳卷烟厂,河南安阳450016

出　　处：《计算机仿真》2025年第1期253-257,共5页Computer Simulation

基　　金：河南中烟工业有限责任公司科技攻关项目(A202046)。

摘　　要：由于拍摄角度、镜头畸变、机械振动等因素,可能会导致图像中的烟梗位置出现偏移或形变,几何失真会直接影响烟梗位置的识别效果,为了解决烟梗按序上料码放位置识别效果不佳的问题,提出了基于改进RANSAC算法的烟梗按序上料码放位置识别方法。通过几何变换方法对CCD相机采集到的烟梗基准图像和烟梗实时上料图像展开几何校正处理,改善位置偏移与形变问题。基于尺度不变特征变换算法选择烟梗图像特征点,并获取特征点描述子,根据特征点描述子实现烟梗基准图像和烟梗实时上料图像的特征点匹配。基于特征点匹配结果,通过改进RANSAC算法完成烟梗按序上料码放位置识别。实验结果表明,所提方法的烟梗按序上料码放位置识别精度更高、效果更好。Due to shooting angle,lens distortion,mechanical vibration and other factors,the position of tobacco stems in the image may be offset or deformed,and geometric distortion will directly affect the recognition effect of tobacco stems.In order to solve the problem of poor recognition effect of tobacco stems in sequence,an improved RANSAC algorithm was proposed.By geometric transformation method,the reference image of tobacco stem and the real-time feeding image of tobacco stem collected by CCD camera were geometrically corrected to improve the position deviation and deformation.Based on the scale-invariant feature transformation algorithm,the feature points of the tobacco stem image were selected,and the feature point descriptors were obtained.According to the feature point descriptors,the feature point matching between the tobacco stem reference image and the tobacco stem real-time feeding image was realized.Based on the matching results of feature points,the position identification of tobacco stems in sequence was completed by improving RANSAC algorithm.The experimental results show that the proposed method has higher accuracy and better effect in identifying the stacking position of tobacco stems in sequence.

关 键 词：SURF算法 几何校正 特征点匹配 烟梗位置识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]