A Simultaneous Multi-view Registration Algorithm for Multiple Range Images

To deal with the problems of large amount of data and complex optimization in global registration, we propose an efficient global registration algorithm. In order to eliminate outliers, we propose a statistic method based on uniform distribution test. In order to accelerate to search for corresponding points, we propose a method to establish a corresponding point cache, which can improve the efficiency of space and time. In order to optimize the error between corresponding points, we propose a method based on pseudoinverse matrix, which is strong in numerical stability. And the algorithm is also fit for parallel computation. It can be proved by experiment that this algorithm can eliminate accumulative error and optimize global error efficiently.