Research On Reconstruction Algorithm Of Point Cloud Obtained By 3D Ultrasound C-scan

Aero-engine blade is one of the key parts of the aero-engine,the safety and the reliability have been affected by the blade's state,so it could make a great sense for meeting the operating requirements of the aircraft,protecting people's property and life safety to realize blade testing efficiently and find the defective blade in time.Ultrasonic nondestructive testing technology has become the important means to test the aero-engine blade on the strength of its high safety,low cost and higher accuracy.Point cloud could be transformed by the data measured by the 3D ultrasound,and with the help of the point cloud processing technology,it could be reconstructed and made transparent.According to the human-machine interaction technology,the 3D model of point cloud could be rotated,scaled and more to help the testing personnel complete the blade detection visually and intuitively.The major elements of this thesis are as follow:(1)The fast radius neighbors search algorithm is presented.The features of the radius neighbors search have been analyzed,and the method of promoting the performance of neighbors has been introduced.An bounding box for the neighbors search scale is established in the Octree from the bottom to the top.The neighbors search is conducted in the bounding box to improve the performance of searching neighbor points,which is accomplished hierarchically by testing the spatial relation between the neighbors search scale and the cell of the bounding box according to the containing test and intersection test under the local coordinate system.(2)The reconstruction algorithm for point cloud obtained by 3D ultrasound C-scan has been presented.To accommodate the particularity of this kind of point cloud,the reconstruction algorithm that using ellipse to approximate the point cloud model is employed to reconstruct this kind of point cloud.The covariance matrix has been built according to the data that acquired by the neighbor points search,the covariance matrix of neighbor points in the x-y plane under the local coordinate system would be gotten by translation,rotation and more transformation operation,next,the ellipse's major and miner axis could be calculated by using the eigenvalues and eigenvectors of this kind of covariance matrix in the local coordinate system,and the center of the ellipse is set,then the model could be reconstruction.The ? value of color channel is set to realize a kind of transparent visualization effect.(3)The the visualization based on the GPU acceleration technology is presented.To get a faster and more real visualization performance,the floating-point computation capability of GPU and CPU are compared,and the GPU acceleration technology is employed to process the reconstructed model.The technical route for GPU acceleration is analyzed,the model trimming,illumination model calculating and color assignment is accomplished in the GPU.The code for these algorithms is written by C++,Open GL and GLSL.The effectiveness of this algorithm that presented in this thesis has been testified by some related experiments.