Utilizing Machine Learning to Improve the Distance Information from Depth Camera

A depth camera provides distance information. However, in the real environment, uncertain measurement conditions may bring incorrect distance information, e.g., environmental conditions, hardware component tolerances, and so on. Thus, we may always obtain unstable and inaccurate information. On the other hand, even sensors with the same specification are used in the experiment, we may obtain different information as well. Therefore, in this work, we intend to solve this issue by incorporating some machine learning approaches in the real environment to improve accuracy and stability. Particularly, we use the concept of machine learning for overall consideration instead of a particular statistics model to evaluate the uncertainty.