Depth Measurement And 3D Topography Reconstruction Based On Monocular Vision

Surface topography and depth measurement has always been an important issue of concern in production and application,which is of great significance to quality assurance,safe operation and reverse engineering.Since of the development of transportation,the detection of tread damage on the train wheels,which plays an important role in ensuring vehicle safety,becomes especially important.However,in practical applications,contact measurement such as artificial visual method and impact load method still dominate.Existing instruments and equipment can only detect three-dimensional topography or surface condition respectively,and an urgent need is a detection method that can simultaneously obtain three-dimensional topography and surface condition of the wheel.In this paper,lots of work were done to search for rapid and accurate method to detect the surface and measure the 3D information.In this paper,the technique of FTP and shape from shading are adopted to achieve 3-D topography reconstruction and defects measurement of tread surface of the train wheel,as well as the problem of high light encountered during image shooting is studied.The main work contents are as follows: 1.To the highlights removal of smooth metal surface,several commonly used methods of image restoration are introduced.Combined with complementary color,color space transformation and double reflection model method the highlight of color metal images are removed for the steel ball and the wheel surface.In order to improve accuracy of the double reflection model method for the black-and-white image,an algorithm of adaptively dividing the highlight area is obtained.2.Based on the principle of defocused fringe projection,experiment is designed to detect the depth of train wheel tread,and using FTP to extract the grating fringe,analyzing the correspondence between height and phase in the experimental structure.In order to extract more information and improve the detection accuracy of FTP in the detection process,three kinds of methods were discussed to recover the background figure from raster image,low-pass filters in the frequency domain,wavelet transform,Butterworth band-stop filter.The reconstruction of the three-dimensional topography of the tested wheel is realized.The test results are analyzed to confine the feasibility of this method.3.Research on the principles and algorithms of shape from shading,and the Tashi algorithm is introduced and applied on the height.measurment of thecomputing image and real image.The fusion result with the depth and the surface image are given.The work and research results in this paper have important reference significance for the research of wheel's defect detection and 3D topography reconstruction.