【作者】 李彬；

【导师】 钟凯；

【作者基本信息】 华中科技大学 ， 材料加工工程， 2023， 硕士

【摘要】 微小零件在机械和电子等领域应用广泛,在加工和组装环节对其进行三维测量能及时发现产品缺陷,有效保证最终质量。基于远心相机的面结构光三维测量技术具有景深大、精度高和非接触等优势,逐渐成为小范围(小于50×50 mm~2)测量的优选方案。然而,其在实际应用时仍存在以下问题:(1)远心相机成像模型特殊,标定时难以有效兼顾灵活性和精度;(2)小范围测量中常因反光、过暗和遮挡导致测量数据完整度低;(3)三维重建计算复杂,较为耗时。为此,本文构建单远心相机双沙姆投影仪结构的远心多目三维测量系统,针对上述问题进行了以下研究:(1)针对远心相机标定难以有效兼顾灵活性和精度的问题,提出基于投影仪补全的远心相机完整参数标定方法。该方法使用系统中已标定的沙姆投影仪进行辅助,并且仅将其用于计算旋转矩阵缺失参数,从而避免带来较大投影仪误差传递。实验结果表明,该方法无需引入系统以外的辅助设备,并且精度优于传统投影仪辅助标定方法,平均重投影误差为2.139μm,兼具高灵活性和高精度。(2)针对反光、过暗和遮挡导致测量数据完整度低的问题,使用多重曝光融合方法提升测量动态范围,并将投影仪当作逆向相机增加测量视角。实验结果表明,以上方法有效提高了电路板等物体的测量数据完整度。同时,针对计算复杂导致重建速度慢的问题,提出基于去畸变查找表的三维重建计算加速方法,该方法通过建立查找表进行投影仪的快速去畸变,以及使用并行计算平台对重建过程进行加速,对比传统CPU平台重建方法,所提方法在不牺牲去畸变精度的前提下三维重建计算效率提升了约60倍。(3)针对微小零件高精度、高完整度和快速的测量需求,集成了远心多目三维测量系统,并基于QT开发了专用测量软件。实验结果表明,本系统的测量精度可达5μm;在选取的典型微小零件的测量上,三维重建时间约0.37 s,有效三维点数量相比于单目单次曝光方法提升了16.53%～35.44%。更多还原

【Abstract】 Small-scale parts are widely used in various fields such as machinery and electronics.Performing 3D measurements on them during the processing and assembly stages can promptly detect product defects and effectively ensure the final quality.The 3D measurement technology of surface structured light based on telecentric cameras has the advantages of large depth of field,high precision,and non-contact,and is gradually becoming the preferred option for small-scale(less than 50×50 mm~2)measurement.However,there are still the following problems in practical applications:(1)The imaging model of the eccentric camera is special,and it is difficult to effectively balance flexibility and precision during calibration;(2)In small-scale measurement,low completeness of measurement data often occurs due to reflection,darkness,and occlusion;(3)The computation of 3D reconstruction is complex and time-consuming.Therefore,this paper constructs a telecentric multi-view 3D measurement system with a single telecentric camera and dual Scheimpflug projectors,and conducts the following research to address the above issues:(1)Aiming at the problem that the calibration of telecentric camera is difficult to effectively balance flexibility and precision,a complete parameter calibration method of telecentric camera based on projector completion is proposed.The method proposed is based on using a Scheimpflug projector that has already been calibrated in the system to calculate the missing rotation matrix parameters,thereby avoiding the introduction of large projector error propagation.The experimental results show that the method does not require the introduction of auxiliary equipment outside the system and has higher precision than traditional projector-assisted calibration methods,with an average reprojection error of2.139μm,while being both highly flexible and precise.(2)Aiming at the problem of low completeness of measurement data caused by reflection,darkness and occlusion,the multi-exposure fusion method is used to improve the dynamic range of measurement,and the projector is used as a reverse camera to increase the measurement angle of view.The experimental results show that the above methods effectively improve the measurement data completeness of objects such as circuit boards.Meanwhile,aiming at the problem of slow reconstruction speed caused by complex calculations,the method for accelerating the 3D reconstruction calculation based on a distortion correction lookup table is proposed.This method establishes a lookup table for fast distortion correction of the projector,and uses a parallel computing platform to accelerate the reconstruction process.Compared with traditional CPU platform reconstruction methods,the proposed method improves the computational efficiency of 3D reconstruction by about 60 times without sacrificing distortion correction precision.(3)To meet the high precision,high completeness,and fast measurement requirements of small-scale parts,a telecentric multi-view 3D measurement system was integrated,and a dedicated measurement software was developed based on QT.The experimental results show that the measurement precision of this system can reach 5μm;in the measurement of selected typical small-scale parts,the 3D reconstruction time is about 0.37 s,and the effective 3D point number is increased by 16.53%to 35.44%compared with the single-camera single-exposure method.更多还原