Research On Robotized Flexible Inspection Technology Based On High Dynamic Range Structured Light Imaging

With the rapid development of aerospace,energy,power technology and other fields,related industries are also increasingly demanding the production quality of large components with complex curved features,such as turbine blades and spacecraft cabin shells.Such parts usually have the characteristics of large size,complex shape,high position accuracy and surface quality requirements,which also put forward higher requirements on the performance of the measuring unit.In order to realize the technical tackling of the above challenges,robotized vision measurement technology is adopted.There are two problems in the application of traditional visual measurement in the industrial manufacturing field: one is that the surface dynamic range of metal parts is large,and images is prone to overexposure,which causes a lot of noise into the measurement result;the second is that the camera resolution is limited,high precision and large field of view is difficult to achieve at the same time.To solve the above two problems,this paper designs a complete measurement scheme:(1)A set of monocular fringe projection structured light equipment is built.Based on the traditional fringe projection measurement principle,fringe generation,multi-step phase-shifting method,phase unwrapping of multi-frequency heterodyne,and phase measurement process based on calibrated coordinate calculation are realized,and a complete monocular fringe projection measurement system is built.In the measurement of non-high light parts,the accuracy of 30 mm diameter measurement reached0.0302 mm,and the accuracy of 100 mm size measurement reached0.0814mm;(2)A measurement method of adaptive projection intensity pixel-level modulation is proposed.The dual-view search method based on epipolar constraint is used to obtain the transformation relationship between camera projectors.The physics-based optical imaging model is used to generate fringe modulation parameters,and the pixel-level modulation fringe pattern algorithm is implemented for fringe projection measurement.Experimental results show that the proposed scheme has a compensation rate of 51.3% in the invalid area of the measurement results in the highlight area,and the measurement accuracy is increased by 77.4%;(3)The probability method of multi-view point cloud splicing based on octree is adopted.The node update function is modeled based on the B-spline error curve,which is used to obtain the node probability update parameter.The multi-view data splicing is realized.Experimental results show that the proposed method fills in the partial lack of a single measurement and the field of view in fusion data is larger than the single frame.In the noise sensitive area,the accuracy of multi-view method is increased by up to 85.32% compared with the single frame measurement accuracy.