Research On The 3D Shape Recovery And Measurement Base On A Single Image

As an important research direction of the engineering and technological field, 3D measurement is playing a more and more important role in all aspects of production and life in society, and has broad application foreground in the field of industry measurement such as product monitor , defect inspection and so on. This article concentrates on research a kind of measurement method-Shape from shading(SFS), which can be applied on non-contact 3D measurement of the product with a single image quickly, ultra-precision, easy operation and low cost can be achieved. Recently, SFS gets rapid development and great improvement in theory research and practice application. Therefore, the research and application of SFS has become a hot topic in the field of 3D vision measurement.At first, a fully survey about 3D measurement research work has been drawn given, and the relative principle and unresolved problem are analyzed. Based on the understanding of the SFS essential, a systematic summarization is made for all current SFS algorithms in the world including Minimization, Propagation, Localization and Linearization approaches. According to the experimental research and theoretical analysis, their range of application and development trend have been, discussed and many problems existed in current SFS algorithms such as low precision, high complicated calculation etc have also been analyzed. This article has introduced a perspective projection model to overcome the defect of low precision, and an improved algorithm has also been developed to solve the model, which could overcome the shortage of the traditional algorithms whose recovery results are too smooth to distortion, because of using the smooth constrains. Meanwhile, it does not need any integrability constraint rigor constraint assumption. To gain a clear image and reduce the influence of noise and distortion, filter technique has been adopted to process the acquired image, and the system calibration technique has been used for the camera calibration and image distortion rectification. Based on the perspective model and algorithms, a 3D vision measurement system on LabVIEW platform with Virtual Instrument technique has been developed in this paper, which can carry on real-time image acquiring, image processing, algorithms computation and analysis. For the sake of advancing the memory spending and operation efficiency, the SFS arithmetic has been realized by transferring the dynamic link libraries file, so that the realization of its software and hardware is economical and high efficiency with simple operation and exploitation time shortening.During the experimental research, a ball has been taken to make precision analysis for all pels, and the method of Minimum Square has been adoped to make ball fitting for the shape error analysis. Then, we have analyzed the major factors such as depth of field, camera lens distortion, light and design of image filter which affect the measurement accuracy. By improving the experiment scheme, we have ulteriorly advanced the 3D measurement precision, and extending this approache to some complexes image measurement.At the end of this paper, the SFS measurement system has been applied for bearing surface defect inspection. The experimental results has shown that the 3D measurement approache developed in this article has not only avoided the rigid constrain condition, but also achieved high precision and efficiency compared with the traditional SFS measurement.