Study On3D Reconstruction Based On Phong Hybrid Reflection Model

In the machine vision, SFS is one of the methods to reconstruct the three-dimensionalsurface of the object. This method can get a three-dimensional shape of the object surfacebased on a single gray image,so its operation is simple and its cost is low. Because of theseadvantages, it can be applied in many fields, such as medical, aerospace, industrialinspection. However, traditional SFS technology based on diffuse reflectance model andknown direction of the light source makes the reconstruction model have a big error,because the reflectance model doesn’t conform to the actual reflection characteritics ofobject surface. And if the direction of the light source is not known, the modelreconstruction can not be achieved. Direct at the above issues, the estimation of the lightdirection and the coefficients of reflectance model has been studied, and the main contentsof the dissertation are as follows:(1) Direct at the direction of light source of the two-dimensional image is unknown,the improved algorithm of Zheng and Chellappa was used to realize the estimation of thelight direction and provide the pre-conditions to achieve the SFS algorithm.(2)The image capture hardware system based on a single image was established, theworking environment of the hardware system was set, the hardware devices in the systemswas selected, the factors which affect the accuracy of the hardware system were analyzed,and the work state of system were given. All of these laid the foundation for the follow-upstudy.(3) The reflection law of the surface was studied, and a method of obtaining thecoefficients of reflectance model according to the extent of area in which the specularreflection can be observed was proposed. What’s more, the samples were made of metaland nylon, the method was validated, Estimation formula of the coefficients of thereflection model based on two materials were got, and a hybrid reflectance modelcorresponding to the actual reflection characteristics of object surface was established. Theaccuracy of the reconstruction of the SFS was improved by the means of the method.(4) Based on the windows7, Open GL and object-oriented programming language Visual C++6.0, the prototype system of the three-dimensional surface reconstruction hasbeen designed and developed based on the mixed reflection and a single image. And theimage processing module, the three-dimensional recovery module, the getting datamodule and the outputting and displaying module have been completed. The correctnessand practicality of the research method of this thesis were verified by living example.