Research On The Measurement Method Of 3D Profile Of High-temperature Object

Forging industry is the base for the development of equipment manufacturing industry.The measurement of the profile of forging surface is one of the most important factors for the development of manufacturing industry.As the country pays more attention to the forging processing quality,technology of 3D measurement has also been developed.Different from measurement of normal temperature object,in measurement of high-temperature object,the high temperature and radiation restrain the conducting of traditional measurement methods.As a non-touch measurement method,visual measurement is featured with high-speed,high-accuracy and convenient operation.Based on the binocular stereo vision measuring principle,the paper conducted a research on the measurement methods of three-dimensional high temperature object.First,the paper designed a proper image collection plan.Use point structured light projection of 450 nm to make characteristics point on the surface of the object.And with the filter of corresponding band to achieve high temperature object image collection.Experiment proves that the plan could manage to collect object image containing cleat characteristics point.Secondly,for the measurement scheme of this paper,research on the NCC stereo matching algorithm based on feature points.Among them,in order to obtain high-quality feature points,several characteristics algorithms have been compared.Harris operator realized the extraction,and construct single-scale feature descriptor of the Harris operator,obtaining original matching point collection according to the comparison of descriptors.Characteristics points focusing on this point is relatively rich in grayscale information,which could improve the accuracy of follow-up grayscale matching.And limiting the search area pixels waiting for being matched within a proper neighbourhood of characteristics points,avoiding lots of useless work.Simultaneously,the paper based on the traditional NCC region matching algorithm,adds pyramid search method.This method can not only improve the computational efficiency while obtaining a large amount of point cloud data,but also further improve the matching accuracy through a search window from large to small.At the end,To obtain relatively complete dense display map,the paper proposed a parallax filling and sub-pixel interpolation method.Parallax filling could fill holes caused by unmatched pixel point in disparity map while sub-pixel interpolation method makes the disparity map smoother,thus to get a mass of point cloud data according to the principle of binocular vision measurement.Based on the measurement plan proposed in this paper,the experiment of the accuracy verification experiment of the measurement system are designed.The experimental results show that the accuracy ofthe measurement system can reach 0.37 mm,which can achieve the measurement purpose better..