An Optimized POSIT Algorithm Based on Mean Convergence

Proportional Orthogonal Projection Iteration (POSIT) is the mainstream iterative algorithm in camera pose estimation. The traditional POSIT algorithm has many iterations, too long operating time, and low algorithm robustness. In response to these problems, it proposes a POSIT optimization algorithm...

Full description

Saved in:
Bibliographic Details
Published in:2021 International Conference on Communications, Information System and Computer Engineering (CISCE) pp. 636 - 640
Main Authors: Ni, Xibing, Zhou, Chunyue, Tian, Hui
Format: Conference Proceeding
Language:English
Published: IEEE 14.05.2021
Subjects:
Filtering algorithms
Filtering theory
machine pose estimation
mean convergence
Pose estimation
POSIT
Robustness
Software
Software algorithms
Time complexity
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Proportional Orthogonal Projection Iteration (POSIT) is the mainstream iterative algorithm in camera pose estimation. The traditional POSIT algorithm has many iterations, too long operating time, and low algorithm robustness. In response to these problems, it proposes a POSIT optimization algorithm based on mean convergence in this paper. The algorithm is based on the traditional POSIT, which the input of each iteration is not the last output, but is determined by the weighted sum of all the previous outputs. The weights are determined by the differences between the outputs and the mean. In order to reduce the number of iterations of the algorithm and the time complexity, the algorithm pre-stores some of the key-value pairs of input and output in the actual scene into the hash table, and uses Bloom filters to save memory space. In theory, the POSIT algorithm can be used down to O(1). Finally, the algorithm is verified by experiments with other existing algorithms in this paper. The results show that the optimized POSIT has significantly improved accuracy and robustness, and its time complexity is lower than that of the traditional POSIT.
DOI:10.1109/CISCE52179.2021.9445890