Research On 3D Shape Measurement Based On Digital Fringe Projection Technology

Three-dimensional(3D)shape measurement based on digital fringe projection,which is with the merits of fast speed,high accuracy and non-contact measurement,has numerous applications including industrial online detection,biomedical engineering,computer science,machine vision and reverse engineering.A typical digital fringe projection system is composed of a projector and a camera.The projector is used to project the fringe patterns onto tested surface and the camera to capture the modulated fringe patterns.The fringe patterns are analyzed to obtain the absolute phase distribution containing the 3D shape information of the object and then the 3D geometry of the tested object can be reconstructed using the internal and external parameters of the calibrated system.Although the digital fringe projection technology has undergone considerable development,and is with higher measurement accuracy and resolution,and faster measurement speed than other 3D shape measurement methods,there are still many factors that restrict its measurement quality.In this thesis,several key issues such as phase unwrapping,system monotonicity and lens distortion correction in digital fringe projection technology are comprehensively analyzed and studied.The main research work in this thesis includes the following five aspects:(1)An improved phase-coding method for absolute phase retrieval based on the path-following algorithm is proposed.Although the traditional phase-coding method for absolute phase retrieval possesses strong robustness,the method leads to unwrapping artifacts near the discontinuity points of 2 as a mass of codewords required.To solve this problem,a novel improved method is proposed in this thesis.The error points around the 2 discontinuities of the recovered absolute phase are located by Canny operator and corrected by the path-following algorithm.Compared with other improved phase-coding methods,only six fringe patterns are required by this method and there is no restriction on the colors of the tested objects.Several experimental results prove the effectiveness of this method.(2)A novel method of geometric dimension assisted absolute phase recovery in 3D shape measurement is presented.In the proposed method,the window Fourier filter-quality guided phase unwrapping(WFF-QGPU)algorithm is used to obtain the relative phase distribution of the tested object from three fringe patterns firstly.Then,the relative phase distribution is converted to absolute phase distribution with the geometric dimension of the tested object as a clue.Compared with temporal phase unwrapping,only three fringe image acquisitions are required to recover the absolute phase,which improves the speed of 3D measurement greatly.As compared with geometric constraint method,there is no limit to the depth range of the tested object in this method.The effectiveness of the proposed method is demonstrated by several experiments.(3)To determine the distance of the virtual plane in the geometric constraint-based absolute phase unwrapping,the monotonicity of the depth z(u ~c,v ~c)with respect to projector pixel coordinateu~p by means of geometric constraint is investigated and a discriminant of?(u ~c,v ~c)determined by the internal and external parameters is presented.As?(u ~c,v ~c)?0 at arbitrary point on the CCD pixel coordinates the minimum depth distance is selected for the virtual plane,and the maximum depth distance is selected as?(u ~c,v ~c)?0.Two structured light systems with different signs of the monotonic discriminant are developed,and the validity and the importance of the monotonic discriminant are experimentally demonstrated.This work complements and improves the geometric constraint-based absolute phase unwrapping algorithm.(4)The geometric constraint-based absolute phase unwrapping method is limited tothe measurement of a certain depth range,making it difficult to measure objects with large depth variation or objects moving in a large depth range.In this thesis a new method to extend the measurement depth limited by geometric constraints for objects with large depth variation is presented.This method mainly includes the following steps:using the watershed image segmentation algorithm to extract the region where the absolute phase distribution obtained by geometric constraints is correct,and making a binary mask of this region;using WFF-QGPU to obtain the relative phase distribution of the tested object;using the binary mask to get the difference between the relative and the absolute phases;converting the relative phase distribution to absolute phase distribution by referring to their phase difference in the mask region.This method has the advantages of no need to increase the number of virtual planes and the image segmentation algorithm being simple.The experimental results prove the effectiveness and correctness of the proposed method.(5)In digital fringe projection system,lens distortion of the camera and projector is the main source of 3D measurement error.Camera calibration has been studied extensively,and the distortion of the camera lens can be well corrected by the distortion parameters.However,the distortion correction of the projector lens remains difficult because of two main reasons.Firstly,the projector cannot be used to directly capture images.Secondly,the correction by the distortion parameters used for the camera lens is not suitable for the projector lens due to its high optical efficiency and optical offset.In this thesis,a new approach of using deep neural networks to address this problem is proposed.The neural network consists of one input layer,five densely connected hidden layers,and one output layer.A ceramic plate with flatness less than0.005 mm is used to acquire the training,validation,and test data sets for the network.It is shown that the measurement accuracy can be enhanced to 0.0165 mm in the RMS value by this technique,which is an improvement of 93.52%.