Research On 3D Reconstruction Technology Based On Focus Variation Microscopic Measurement

At present,in the field of micro nano measurement,it is still a challenging task to measure the three-dimensional topography of complex microstructure.It is a little difficult to measure the slope surface with a big gradient with now existing non-contact measurement method although the non-contact methods may have a high vertical resolution.In recent years,the focus variation has been expected to break this dilemma and realize the three-dimensional surface measurement of the microscopic complex structure.A focus variation system includes optical imaging system and motion control system.In the optical imaging system,it must be stressed that the objective lens with ultra short depth of field are chosen and they play a crucial role.And the motion control system can change the distance between the measured object and objective lens to achieve the vertical stratified imaging of the measured object.So image sequences can be collected at different working distances and the measurement can be completed by finding the best focusing position.The focus variation method can measure the complex micro structure with large depth to width ratio and large inclination angle,and can measure a cross section at a time.It has the advantages of fast measuring speed and high measuring precision?According to the measuring principle of focus variation,the experimental measurement platform was built.Two dimensional image sequences with different focusing positions were collected by changing the distance between the measured object and the imaging objective lens.Aiming at the phenomenon of pixel drift during the acquisition of the twodimensional image sequences,a digital image correlation algorithm by sub area matching was compiled to do the displacement correction in subsequent focus evaluation algorithm.By analyzing and comparing the focus evaluation curves obtained by various focus evaluation functions,the focus evaluation function with single peak and strong robustness was chosen.In order to get the depth value of the pixel point,such search algorithm was proposed that firstly,the maximum value in a large range must be found and then the extreme value of the small neighborhood near the maximum value can be got by Gaussian fitting algorithm.The height of all the measured points can be obtained by the above operation.Then,the gray fusion image and color fusion image were obtained by the image fusion algorithm based on the region feature in the space domain,and the image fusion results of the frequency domain were obtained by using the wavelet transform.Finally,the three dimensional points were imported into the software Cloud Compare,and the three-dimensional figure of the measured object could be shown.Besides,the gauge blocks were used to verify the accuracy of the measurement and good results were achieved.