Research On White Light Interference Three-dimensional Topography Stitching Technology

White light micro-interferometry technology is an important branch in the field of interferometry.Compared with monochromatic light interferometry,it can detect the large deviation and step shape of real samples.It is used in the fields of micro-electromechanical,micro-mechanical,micro-optics,etc.Advanced manufacturing has extensive and important application value.While the white light interferometer achieves high resolution through the micro interference objective lens,the horizontal measurement range is limited by the objective lens field of view.Therefore,realizing large-scale and high-resolution shape detection in the horizontal direction has always been the focus and difficulty in the field of white light interference technology.Aiming at the problem that the noise area of the three-dimensional shape affects the stitching accuracy,a weighted reconstruction algorithm based on the quality of the interference fringe is proposed to realize the reconstruction of the noise area.In order to study the technology of 3D shape splicing based on the overlap area,this paper compares and analyzes various algorithms in principle and performs simulation verification.In the in-plane registration stage,based on the different surface features of the two types of samples,the step type and the near-plane type,combined with the feature detection method of the Canny operator and the DoG operator,an edge detection method is proposed.An adaptive template matching method based on splicing direction can solve the contradiction between template size and registration accuracy.In the phase of out-of-plane error correction and shape fusion,the common out-of-plane error correction algorithm and shape fusion algorithm are compared and analyzed,and finally the better global stitching algorithm and linear weighted fusion algorithm are adopted.Finally,on the self-developed white light micro-interferometer in the laboratory,I do the splicing experiments and make error analysis.The experimental results show that the horizontal splicing accuracy of the step-type sample splicing experiment is 0.36 ?m,the vertical splicing accuracy is 3.58 nm,the horizontal splicing accuracy of the near-planar sample splicing experiment is 1.414 ?m,the Rq value error is 0.72 nm.It is verified that the algorithm in this paper can achieve large-scale and high-resolution shape detection of step samples and nearplanar samples.