Research On The Key Technology In Phase Measuring Profilometry Based On The Wavelet Transform

Three-dimensional reconstruction of object is a fast and accurate measurement technique to obtain the object's three-dimensional surface information,which has been widely used in three-dimensional printing,object recognition,cultural relics protection,the printed circuit board,a number of products in the field of manufacturing,modern medicine and so on.The phase measuring profilometry as its non-contact measurement,the measuring object without injury,and measurement speed,high precision,wide measuring scene,has become a hot research direction of three-dimensional reconstruction technique.Phase measuring profilometry's main principle is based on phase modulation and optical triangulation.Extract the phase shift(relative phase,also named wrapped phase)caused by the height of object from two-dimensional modulated image,and unwrap the extracted modulated phase,then we can get the three-dimensional data of the measured object through by the relationship between the phase and height in the optical measurement system.Phase measuring profilometry contains lots of key technical points,like extracting relative phase(phase extraction),relative phase unwrapping(unwrap phase),coordinate data conversion,etc.Although the phase measuring profilometry has its own advantages,but also there are lots of problems need to improve and research in many aspects: 1)Because of the height transform varying magnitudes,and is affected by the air dust,instrument jitter,the phase information will be larger error when extracted from the deformation grating;2)In phase unwrapping process,it is also easy to result in phase unwrap error because of invalid,noise and shadows and other factors.finally,we can't obtain accurate real phase.In this paper,the phase measuring profilometry based on wavelet transform(wavelet transform profilometry),From the two key technical points: phase extracted,phase unwrapping,the main research work and improvement made as follows:(1)The traditional wavelet transform profilometry is sensitive to low noise,which resulting in the problem of inaccurate phase extraction,an improved wavelet ridge phase extraction algorithm based on cost function is proposed.To overcome the lost ridge points during extracting candidate wavelet ridge points,The algorithm extract the maxima points by the wavelet transform coefficients modulus,then get the inflection points which were easily ignored through by the phase information,so the complete candidate wavelet ridge points were obtained;Then,improve the weight factors in raw cost function through by introducing the logarithm logistic mode,which balanced the importance of the two indicators in cost function,then obtain more reasonable cost estimate.Finally,quickly and accurately identified the optimal cost function line from the ridge candidate ridge points by dynamic programming ideas which is in cost function algorithm,and phases on ridge line were extracted.Comparative experimental results show that: In low level noise,our algorithm can effectively suppress the influence of noise,precisely position wavelet ridge,extract more accurate phase information.(2)The wavelet transform profilometry exist problems like phase discontinuity,the residual factor,shaded areas,which leading wrapped phase can't be quickly and accurately unwrapped,a new phase unwrapping algorithm which is suitable to wavelet transform profilometry and could effectively suppress noise is proposed.The algorithm construct a new quality map named "wavelet ridge amplitude coefficient of variance derivative",which though by wavelet ridge amplitude coefficient matrix and the variance of the derivative thought;and use the map to guide phase unwrapping,then could obtain a complete real phase which include the height information of the object.The proposed quality map is simple to construct,In phase unwrapping process,not only can get a more accurate real phase values,but also effectively suppress the influence of noise,then could reconstruct a high accuracy three-dimensional model of an object.(3)Integrate the two above proposed algorithms to achieve the simulation,and implementation details and results are given.