In most conventional algorithms of registering multiple range images, the pose parameters are estimated by using the distance sum between closest point pairs as the objective function. These algorithms have the proble...In most conventional algorithms of registering multiple range images, the pose parameters are estimated by using the distance sum between closest point pairs as the objective function. These algorithms have the problems of inexact point correspondence, registration accuracy, and sensitivity to initial registration parameters. Due to the scanner settings, scanner distance, and surface slopes, two or more 3D data sets are unlikely to be acquired such that the 3D data points exactly correspond, and also each point in the data set may represent different surface areas. This paper proposes a novel registration algorithm based on a distance metric of surface-to-surface. The algorithm uses triangle meshes to represent the surfaces. Based on surface sampling and the point-to-surface distances, the integration calculation of the mean distance between surfaces is derived and reduced to a simple formula. The method was tested on synthetic and real range images.展开更多
基金This work was supported by the National Natural Science Foundation of China(Grant No.60275001).
文摘In most conventional algorithms of registering multiple range images, the pose parameters are estimated by using the distance sum between closest point pairs as the objective function. These algorithms have the problems of inexact point correspondence, registration accuracy, and sensitivity to initial registration parameters. Due to the scanner settings, scanner distance, and surface slopes, two or more 3D data sets are unlikely to be acquired such that the 3D data points exactly correspond, and also each point in the data set may represent different surface areas. This paper proposes a novel registration algorithm based on a distance metric of surface-to-surface. The algorithm uses triangle meshes to represent the surfaces. Based on surface sampling and the point-to-surface distances, the integration calculation of the mean distance between surfaces is derived and reduced to a simple formula. The method was tested on synthetic and real range images.
