摘要
In this paper,a new integrated registration algorithm based on neural network and ICP（iterative closest point） algorithm is presented.A coarse registration process is implemented with neural network,and then further optimized by ICP algorithm.The corresponding point pairs are found according to the target points＇ curvature and color information.Mahalanobis distance,which reflects the scattering degree of point data,is employed to define the closest distance and the closest points.The results of the experiment show that our algorithm has better feasibility,validity and efficiency than the traditional ICP method.
In this paper,a new integrated registration algorithm based on neural network and ICP（iterative closest point） algorithm is presented.A coarse registration process is implemented with neural network,and then further optimized by ICP algorithm.The corresponding point pairs are found according to the target points＇ curvature and color information.Mahalanobis distance,which reflects the scattering degree of point data,is employed to define the closest distance and the closest points.The results of the experiment show that our algorithm has better feasibility,validity and efficiency than the traditional ICP method.
基金
Supported by the National Natural Science Foundation of China (69775022)
