| With the development of high-precision manufacturing,quality inspection of workpiece surfaces has gradually become a research hotspot.Currently,widely used visual measurement technologies are generally based on the principle of triangulation,which focuses on obtaining height information of objects and cannot capture local surface details,resulting in threedimensional models that are not suitable for defect detection.Traditional defect detection techniques based on 2D images lack depth information.Therefore,this thesis proposes a seamless fusion method of surface multi-projector structured light and photometric stereo.The surface structured light system generally consists of a projector and a camera,where the projector projects a specific pattern of raster graphics onto the surface of the object being measured,and the camera captures the image of the object’s surface that is projected.By using algorithms to adjust the captured images,the depth of the object’s surface is obtained,and the surface depth is determined by triangulation.Although this method is easy to operate,the continuity of the data is not high.Photometric stereo uses six light sources to reconstruct the surface with good smoothness and continuity.The fusion measurement results obtained can ensure the measurement accuracy of macroscopic dimensions and improve the precision of detail representation.The main content and innovation points of this research are as follows:(1)The basic principles of photometric stereo are introduced,including photometry,radiometry,and commonly used reflection models such as bidirectional reflectance distribution function.(2)The calibration of photometric stereo light sources is implemented,the normal field of photometric stereo is constructed,and the depth information is solved using the gradient integration algorithm.(3)The method for constructing the normal field of the stripe projection measurement is introduced,and the high-frequency and low-frequency error distribution characteristics of the normal and depth fields of structured light and photometric stereo are analyzed.(4)The surface energy model is established based on the fast convolution method for normal field smoothing of stripe projection measurement,and the low-frequency information stability of the normal field of stripe projection measurement and the low-frequency deviation of the normal field of photometric stereo are obtained.(5)The high-frequency detail transfer model of the photometric stereo normal field is constructed,and the uniform high-frequency correction of the normal field of the stripe projection measurement is achieved using the transfer field.(6)The method of establishing the surface energy model based on the joint modulation of position error and normal error is studied,and the super-constraint sparse linear system is constructed to achieve efficient optimal mesh surface calculation. |
