Kinfu Key Technology Research And Improvement Of The Algorithm

With the development of computer hardware,computer vision field with more powerful background support,two-dimensional images already can’t satisfy the needs of research,attaches great importance to the 3d scene is imperative.At present there are signs extraction algorithms for 3 d reconstruction algorithm,the reconstruction of the block matching,and the point clouds reconstruction,however,these technologies have some limitations in terms of reconstruction.Using the Kinect sensor to get the depth image,the data alignment of the multiple perspectives,and then the 3d reconstruction of a point cloud fusion.However,the development of this technique is not very mature,and in the reconstruction image,the image is empty,the efficiency is slow,the effect frame number is low.To improve the problem of the Kinect’s remodeling,the Kinect will be able to do better in many areas.In this paper,the key algorithms in Kinfu algorithm was improved,the key algorithm for experiments and theoretical analysis,the realization of effective,rapid and accurate 3 d scene reconstruction.The key problem in the Kinfu algorithm is the processing of the cloud preprocessing,and when dealing with the cloud,the algorithm is complex and computationally intensive.The point cloud obtained by the Kinect is arranged in an orderly fashion according to the matrix,in this paper,we use the traversal points to make the triangle surface and quickly triangulate.Registration phase,the key point of point cloud are extracted,the use of key points for registration,on the premise of guarantee registration results,speed up the registration,improve the algorithm running time.In the reconstruction phase,the color image and the depth image pixels are combined.Because Kinfu single only to the depth image,color images are not used,this article on the basis of predecessors’ use of color image edge,the use of color images of each pixel to match the depth image pixels,make the reconstructed image has good integrity,fill the empty image,image reconstruction is more complete.