Acquisition And Processing Based On The Wavelet Transform For 3D Data Of Object Surface

Three-dimensional measurement technology,as a technology that can get threedimensional information and three-dimensional model of the measured object's surface,has been widely applied in three-dimensional printing,machine vision,virtual reality and other fields.Wavelet transform profilometry is a three-dimensional measurement technology based on optics.It has become a research hotspot in three-dimensional measurement technology because of its advantages such as non-contact,strong noise immunity,high reconstruction accuracy,only one image is needed to obtain three-dimensional information and so on.The basic principle of wavelet transform profilometry is to get the deformation fringe pattern by projecting grating to the surface of the measured object,and extract the modulated phase caused by the height change of the measured object from the deformation fringe pattern,so as to get the height information of the measured object.The main steps of acquiring three-dimensional information and three-dimensional model of the measured object by using wavelet transform profilometry include system parameter calibration,phase extraction,phase unwrapping,phase height mapping,point cloud denoising,point cloud simplification,point cloud registration,surface reconstruction and so on.In view of the two key technical points of phase extraction and point cloud registration,the main work and research content of this paper are as follows:(1)Aiming at the problem that the wavelet ridge is easily affected by noise,which can result in the inaccuracy of wavelet ridge extraction and large phase error,a wavelet ridge extraction algorithm combined with spatial noise information is proposed in this paper.In the algorithm,the information of the wavelet transform coefficient amplitude and the phase is combined to obtain more perfect candidate ridge points;the cost function is established by using the information of the wavelet transform coefficient amplitude and the scale parameter curve gradient;the Feature Scaling method is used to adjust wavelet transform coefficient amplitude and scale parameter curve gradient to balance the weight of the two indexes;increases along with the noise,the wavelet ridge is more affected by the noise,combined with the spatial noise information to compensate the adjusted scale parameter curve gradient,increase the weight of scale parameter curve gradient,make the wavelet ridge more smooth.The experiment shows that the algorithm has good anti-noise ability and robustness,and can extract the wavelet ridge accurately.(2)Aiming at the problem of low precision and slow speed in the 3D point cloud registration process,an ICP algorithm based on multi-resolution registration point is proposed in this paper.First of all,the concept of multi-resolution is introduced by this algorithm,the points in the point cloud are graded by using the average value information of the normal vector angle,and the points of each level are sampled different at different resolutions,the use of low-resolution matching point to quickly complete the registration,high-resolution matching points to improve the registration accuracy;And then the concept of matching degree is introduced to find a more suitable matching point;KD-tree is used to speed up the searching between points and calculate the curvature information of the corresponding points according to the computation needs,so as to further improve the registration efficiency.The comparison experiment shows that the algorithm has a great improvement in precision and speed,and with the increase of point cloud scale,the speed increase trend is increasing progressively.