Statistical Robustness Analysis of Random Sampling Consensus Method

One of the most vocalized applications of computational interactivity today stem from our biological sense of perception, both in its promise for automation and heeding of its still prevalent weaknesses. Computer vision, as it is known, is a rapidly growing sub-field of computer science that creates use out of visual input utilizing various vision models and algorithms. Naturally these models and algorithms vary widely in terms of correctness, robustness, and degeneracy, especially when operating under disparate environments and conditions. Many publications explore the goal of developing new and robust vision models or algorithms, but less so explore the comparisons between those that already exist. The purpose of this paper is to detail the performance of Visual SLAM with other modern computer vision models (such as PTAM, ORB-SLAM, DSO, LSD, etc.) to produce a standard by which full comparisons may be drawn for both disparate environmental and conditional datasets. It is hoped that this paper will inform others in academia of the current state of computer vision models and help determine when the use of one model should be preferred over another given a certain environment and/or operating condition