3D deformation measurement of rotating blades based on concentric ring calibration and GPU-SIFT feature point searching

Abstract

Image grayscale matching and camera calibration are two key technologies for measuring the three-dimensional deformation of rotating blades. To address the issues of low accuracy in extracting corresponding points caused by unclear grayscale boundaries of traditional chessboard calibration patterns and the high computational complexity of image matching algorithms in high-speed and high-resolution scenarios, this paper proposes a method for measuring the three-dimensional deformation of rotating blades based on concentric circle calibration and GPU-SIFT feature point search. Firstly, a concentric circle calibration method based on arc segment combination for extracting circle center is proposed. By fitting the circular calibration pattern with extracted marker points, errors introduced by unclear grayscale boundaries of chessboard patterns are avoided, thereby improving the accuracy of calibration parameters. Secondly, an image matching algorithm based on GPU-SIFT feature point search is presented. By optimizing the storage location of scale space process data, the performance of the SIFT algorithm is maximized, leading to improved image matching speed. Lastly, the calculation of displacement and strain fields is achieved using the P3P theory and local least squares fitting method, respectively. The experimental results show that the proposed system and method for 3D deformation measurement achieve a strain measurement accuracy of 80 με. Moreover, they successfully measure the surface deformation of turbofan blades under the conditions of 3000 rpm and 6000 rpm. Through this method, it can improve the deformation measurement capability of large engine manufacturers for high-speed rotating blades and provide reliable technical support.